## The problem

Choosing a local GGUF model for a 16 GB machine is mostly vibes. People recommend "just use Q4_K_M" or "try Qwen 7B" but nobody systematically tests what actually works best **on this specific hardware class** across throughput, latency, and quality simultaneously. I wanted hard numbers.

## What I tested

- **331 GGUF models** across 48+ model families (Qwen, Nemotron, Devstral, GLM, LFM2, GPT-OSS, Ling, and more)
- Quantizations from IQ1_M to F16/BF16
- Sizes from 0.8B to 35B parameters
- 37 waves of automated benchmarking

## How I tested

Each model goes through this automated pipeline:

1. **Download** from Hugging Face
2. **Throughput benchmark** — `llama-server` perf sweep at 1024 and 4096 context, concurrency 1 and 2, measuring TTFT (time to first token), tokens/sec, end-to-end latency, and time per output token — all at p50
3. **Quality evaluation** in two lanes:
   - **Reasoning** — GSM8K (20 math problems) + MMLU (60 questions) via `/v1/chat/completions`
   - **Non-reasoning** — same datasets via `/v1/completions` (no chat template)
4. **Upload** tested GGUF + benchmark artifacts to Hugging Face
5. **Delete** local GGUF to free disk for the next wave
6. **Repeat** — fully automated, recoverable from crashes, resumable from state file

The whole thing runs unattended. One Mac Mini. No cloud. The pipeline script, benchmark harness, and all data are in the repo (link at the bottom).

## Key findings

### 31 out of 331 models are degenerate on 16 GB

Models with TTFT > 10 seconds or < 0.1 tokens/sec are memory-thrashing — they technically run but are unusable. This includes all 27B+ dense models and some poorly quantized 24B models. **If a model's weights + KV cache exceed ~14 GB, performance falls off a cliff on this hardware.**

| Model | TTFT (1k c1) | TPS (1k c1) | Note |
|-------|-------------|-------------|------|
| Qwen3.5-27B-heretic-v2-Q4_K_S.gguf | 97.30s | 0.0069 | TTFT 97.3s |
| Qwen3.5-27B-Q4_K_M.gguf | 66.93s | 0.0077 | TTFT 66.9s |
| Devstral-Small-2-24B-Instruct-2512-Q4_K_M.gguf | 65.32s | 0.0198 | TTFT 65.3s |
| Qwen3.5-27B-Q4_0.gguf | 60.89s | 0.0090 | TTFT 60.9s |
| qwen2.5-coder-14b-instruct-q8_0.gguf | 58.85s | 0.0157 | TTFT 58.9s |
| Qwen3.5-27B-IQ4_NL.gguf | 52.41s | 0.0090 | TTFT 52.4s |
| Qwen_Qwen3.5-27B-IQ4_NL.gguf | 46.05s | 0.0122 | TTFT 46.1s |
| Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf | 42.65s | 0.0256 | TTFT 42.7s |
| Qwen3.5-27B-Q4_K_S.gguf | 34.83s | 0.0156 | TTFT 34.8s |
| ...and 22 more (see ANALYSIS.md for full list) |

### The Pareto frontier has 11 models

After filtering out degenerate models, I computed a composite quality score (average of GSM8K + MMLU across both reasoning and non-reasoning) and plotted it against throughput. Only **11 models** are Pareto-optimal — meaning no other model beats them on BOTH speed and quality:

| Model | Source | TPS (avg) | Quality | Reasoning GSM8K | Reasoning MMLU | Non-Reasoning GSM8K | Non-Reasoning MMLU |
|-------|--------|-----------|---------|-----------------|----------------|--------------------|--------------------|
| Huihui-Ling-mini-2.0-abliterated.i1-Q4_K_S.gguf | mradermacher | 50.34 | 24.2 | 5.00% | 41.67% | 0.00% | 50.00% |
| inclusionAI_Ling-mini-2.0-IQ4_NL.gguf | bartowski | 49.78 | 25.8 | 5.00% | 50.00% | 5.00% | 43.33% |
| inclusionAI_Ling-mini-2.0-Q3_K_L.gguf | bartowski | 46.25 | 26.2 | 15.00% | 30.00% | 0.00% | 60.00% |
| Huihui-Ling-mini-2.0-abliterated.i1-Q3_K_L.gguf | mradermacher | 46.02 | 28.3 | 10.00% | 41.67% | 0.00% | 61.67% |
| Ling-Coder-lite.i1-IQ4_NL.gguf | mradermacher | 24.31 | 29.2 | 0.00% | 48.33% | 25.00% | 43.33% |
| Ling-Coder-lite.i1-Q4_0.gguf | mradermacher | 23.58 | 31.3 | 5.00% | 48.33% | 20.00% | 51.67% |
| LFM2-8B-A1B-Q5_K_M.gguf | unsloth | 19.68 | 44.6 | 50.00% | 48.33% | 40.00% | 40.00% |
| LFM2-8B-A1B-UD-Q5_K_XL.gguf | unsloth | 18.85 | 44.6 | 50.00% | 48.33% | 40.00% | 40.00% |
| LFM2-8B-A1B-Q8_0.gguf | unsloth | 15.08 | 46.2 | 65.00% | 46.67% | 25.00% | 48.33% |
| LFM2-8B-A1B-UD-Q8_K_XL.gguf | unsloth | 14.88 | 47.9 | 55.00% | 46.67% | 40.00% | 50.00% |
| LFM2-8B-A1B-UD-Q6_K_XL.gguf | unsloth | 13.95 | 50.4 | 60.00% | 50.00% | 35.00% | 56.67% |

Every other model in the 331 is strictly dominated by one of these eleven. The frontier spans from pure speed (Ling-mini at 50 tok/s) to pure quality (LFM2-8B-A1B at 50.4 composite).

### Context scaling is surprisingly flat at 4k

The median TPS ratio (4096 / 1024 context) across viable models is **1.0** — most models show zero degradation going from 1k to 4k context on this hardware. Some actually speed up (the LFM2 models gain 15-30% at 4k). This suggests the memory bandwidth bottleneck hasn't kicked in yet at 4k context, which is promising. The open question is where the cliff is at 8k, 16k, 32k — that's the next test.

### Concurrency scaling is poor (memory bandwidth bound)

At concurrency 2, models average only **0.57x** the throughput of concurrency 1 (ideal would be 2.0x). This confirms the Mac Mini is memory-bandwidth limited — two concurrent requests fight for the same unified memory bus. Running a single request at a time is the right strategy on this hardware.

## Top 3 recommendations for 16 GB Mac Mini

### 1. LFM2-8B-A1B-UD-Q6_K_XL.gguf (unsloth) — Quality champion

- **Quality: 50.4** composite (60% reasoning GSM8K, 50% reasoning MMLU, 35% non-reasoning GSM8K, 56.67% non-reasoning MMLU)
- 13.95 tok/s avg, 0.48s TTFT, 0.90s E2E
- Highest quality of any viable model across all 331 tested. Q6_K gives a good middle ground on KV cache pressure. Pareto optimal.

### 2. LFM2-8B-A1B-Q5_K_M.gguf (unsloth) — Speed champion (among quality models)

- **19.68 tok/s avg** (20 tok/s at 1k context!)
- Quality: 44.6 composite (50% reasoning GSM8K, 48.33% reasoning MMLU, 40% non-reasoning GSM8K, 40% non-reasoning MMLU)
- Smallest LFM2 quantization = most headroom for longer contexts and KV cache. Pareto optimal.

### 3. LFM2-8B-A1B-UD-Q8_K_XL.gguf (unsloth) — Balanced pick

- Quality: 47.9 composite (55% reasoning GSM8K, 46.67% reasoning MMLU, 40% non-reasoning GSM8K, 50% non-reasoning MMLU)
- 14.88 tok/s avg — strong speed with near-top quality
- Pareto optimal

### Honorable mention: Ling-mini for pure speed

- Ling-mini models (MoE) achieve **40-50 tok/s** — 3x faster than LFM2 — but with lower quality (22-28 composite). If you need raw speed over quality, `Huihui-Ling-mini-2.0-abliterated.i1-Q4_K_S.gguf` at 50.3 tok/s is the speed king.

### Why LFM2 dominates

LFM2-8B-A1B is an 8B **mixture-of-experts (A1B = 1B active parameters)** model. It's architecturally ideal for memory-limited hardware: only ~1B params are active per token, so the memory bandwidth pressure per token is much lower than a dense 8B model. This is why it achieves 12-14 tok/s while dense 8B models top out at 5-7 tok/s on the same hardware.



## Ling Family Analysis (W23-W27)

This slice includes all `Ling-mini-2.0*` and `Ling-Coder-lite*` runs across:
- `mradermacher/Ling-mini-2.0-GGUF`
- `bartowski/inclusionAI_Ling-mini-2.0-GGUF`
- `inclusionAI/Ling-mini-2.0-GGUF`
- `mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF`
- `mradermacher/Ling-Coder-lite-i1-GGUF`

Coverage:
- Total Ling-family models evaluated: `48`
- Models with complete perf+quality metrics: `47`
- Models with missing metrics: `1` (`Huihui-Ling-mini-2.0-abliterated.i1-Q6_K.gguf`)

Distribution summary (47 complete models):
- Composite quality (avg of R-GSM8K, R-MMLU, NR-GSM8K, NR-MMLU): mean `24.54`, median `24.58`, range `14.58` to `31.25`
- Throughput (`TPS_p50_avg`): mean `27.43`, median `31.44`, range `10.78` to `40.33`
- TTFT (`TTFT_p50_1k_c1`): mean `0.77s`, median `0.48s`, range `0.26s` to `5.28s`

Observed tradeoff:
- Correlation between throughput and composite quality is moderately negative (`r = -0.454`), so the fastest variants are not usually the highest-quality variants.

### Pareto-optimal Ling models (quality vs speed)

Pareto definition used: maximize both `composite quality` and `TPS_p50_avg`.

| Model | Source | Composite | TPS_p50_avg | TTFT_p50_1k_c1 |
|---|---|---:|---:|---:|
| Ling-Coder-lite.i1-Q4_0.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 31.25 | 19.52 | 0.801s |
| Ling-Coder-lite.i1-IQ4_NL.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 29.16 | 20.82 | 0.652s |
| Huihui-Ling-mini-2.0-abliterated.i1-Q3_K_L.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 28.34 | 36.59 | 0.434s |
| inclusionAI_Ling-mini-2.0-Q3_K_XL.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 25.84 | 36.70 | 0.336s |
| inclusionAI_Ling-mini-2.0-IQ4_NL.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 25.83 | 40.33 | 0.259s |

Interpretation:
- If quality is primary, the `Ling-Coder-lite` branch dominates the top quality ranks.
- If throughput is primary, the `inclusionAI_Ling-mini-2.0` and `Huihui-Ling-mini` branches dominate.
- `inclusionAI_Ling-mini-2.0-IQ4_NL.gguf` is the speed anchor (highest average TPS), while `Ling-Coder-lite.i1-Q4_0.gguf` is the quality anchor.

### Family-level comparison

Average across each source family:

| Source family | Models | Mean Composite | Mean TPS_p50_avg |
|---|---:|---:|---:|
| mradermacher/Ling-Coder-lite-i1-GGUF | 12 | 28.09 | 15.15 |
| mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 12 | 23.98 | 33.03 |
| bartowski/inclusionAI_Ling-mini-2.0-GGUF | 15 | 23.86 | 31.50 |
| mradermacher/Ling-mini-2.0-GGUF | 6 | 22.57 | 31.49 |
| inclusionAI/Ling-mini-2.0-GGUF | 3 | 19.72 | 27.55 |

## Consolidated benchmark table (all 331 models)

Sorted by composite quality score (average of Reasoning GSM8K, Reasoning MMLU, Non-Reasoning GSM8K, Non-Reasoning MMLU). Models marked with \* are degenerate (TTFT > 10s or TPS < 0.1 at 1k context).

| # | Model | Source | TPS (1k c1) | TPS (4k c1) | TTFT (s) | E2E (s) | R-GSM8K | R-MMLU | NR-GSM8K | NR-MMLU | Composite |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | LFM2-8B-A1B-UD-Q6_K_XL.gguf | unsloth/LFM2-8B-A1B-GGUF | 13.4388 | 14.4606 | 0.4840 | 0.8970 | 60.00% | 50.00% | 35.00% | 56.67% | 50.4 |
| 2 | LFM2-8B-A1B-UD-Q8_K_XL.gguf | unsloth/LFM2-8B-A1B-GGUF | 14.9344 | 14.8275 | 0.4423 | 0.9080 | 55.00% | 46.67% | 40.00% | 50.00% | 47.9 |
| 3 | LFM2-8B-A1B-Q6_K.gguf | unsloth/LFM2-8B-A1B-GGUF | 12.0456 | 15.9689 | 0.4668 | 0.8716 | 60.00% | 45.00% | 35.00% | 48.33% | 47.1 |
| 4 | LFM2-8B-A1B-Q8_0.gguf | unsloth/LFM2-8B-A1B-GGUF | 14.0586 | 16.1106 | 0.5221 | 0.9782 | 65.00% | 46.67% | 25.00% | 48.33% | 46.2 |
| 5 | MechaEpstein-8000M-Q4_K_M.gguf | ortegaalfredo/MechaEpstein-8000-GGUF | 4.8439 | 6.1937 | 0.6157 | 1.3745 | 70.00% | 58.33% | 5.00% | 51.67% | 46.2 |
| 6 | LFM2-8B-A1B-Q5_K_M.gguf | unsloth/LFM2-8B-A1B-GGUF | 20.0332 | 19.3233 | 0.3894 | 0.6990 | 50.00% | 48.33% | 40.00% | 40.00% | 44.6 |
| 7 | LFM2-8B-A1B-UD-Q5_K_XL.gguf | unsloth/LFM2-8B-A1B-GGUF | 17.5638 | 20.1436 | 0.4264 | 0.7452 | 50.00% | 48.33% | 40.00% | 40.00% | 44.6 |
| 8 | mistralai_Devstral-Small-2-24B-Instruct-2512-Q3_K_S.gguf | bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF | 2.3405 | 2.1575 | 2.7171 | 7.4623 | 0.00% | 58.33% | 25.00% | 70.00% | 38.3 |
| 9 | Qwen3-14B-GPT-5.2-High-Reasoning-Distill.q3_k_m.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 3.1737 | 3.1659 | 1.2783 | 3.5126 | 15.00% | 33.33% | 25.00% | 66.67% | 35.0 |
| 10 | Qwen3-14B-GPT-5.2-High-Reasoning-Distill.q4_k_m.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 4.5345 | 4.4505 | 0.9060 | 2.5380 | 10.00% | 43.33% | 20.00% | 61.67% | 33.8 |
| 11 | Qwen3-14B-GPT-5.2-Distill.q3_k_m.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 4.1123 | 4.1214 | 1.2819 | 3.5259 | 15.00% | 48.33% | 15.00% | 53.33% | 32.9 |
| 12 | LFM2-24B-A2B.IQ4_XS.gguf | mradermacher/LFM2-24B-A2B-GGUF | 15.2790 | 14.3557 | 0.5005 | 0.9015 | 10.00% | 60.00% | 15.00% | 46.67% | 32.9 |
| 13 | LFM2-24B-A2B.Q4_K_S.gguf | mradermacher/LFM2-24B-A2B-GGUF | 3.0243 | 2.4577 | 3.0913 | 5.6132 | 15.00% | 58.33% | 20.00% | 33.33% | 31.7 |
| 14 | mistralai_Devstral-Small-2-24B-Instruct-2512-Q3_K_M.gguf | bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF | 2.4334 | 1.6735 | 2.4643 | 6.8444 | 0.00% | 45.00% | 15.00% | 66.67% | 31.7 |
| 15 | Ling-Coder-lite.i1-Q4_0.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 23.1897 | 23.9760 | 0.8006 | 1.2442 | 5.00% | 48.33% | 20.00% | 51.67% | 31.3 |
| 16 | LFM2-24B-A2B.Q3_K_L.gguf | mradermacher/LFM2-24B-A2B-GGUF | 10.3176 | 11.3589 | 0.7629 | 1.2837 | 10.00% | 55.00% | 20.00% | 36.67% | 30.4 |
| 17 | Ling-Coder-lite.i1-Q3_K_L.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 20.3339 | 23.0096 | 0.4412 | 0.9485 | 5.00% | 43.33% | 30.00% | 41.67% | 30.0 |
| 18 | Ling-Coder-lite.i1-Q4_1.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 21.5602 | 23.5536 | 0.4666 | 0.9145 | 0.00% | 48.33% | 20.00% | 51.67% | 30.0 |
| 19 | LFM2-24B-A2B-Q4_0.gguf | LiquidAI/LFM2-24B-A2B-GGUF | 4.2990 | 3.4663 | 1.5770 | 3.1035 | 15.00% | 61.67% | 15.00% | 28.33% | 30.0 |
| 20 | Ling-Coder-lite.i1-IQ3_XS.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 14.9873 | 15.1814 | 0.6844 | 1.4232 | 5.00% | 43.33% | 25.00% | 45.00% | 29.6 |
| 21 | qwen2.5-coder-14b-instruct-q6_k.gguf | Qwen/Qwen2.5-Coder-14B-Instruct-GGUF | 2.6008 | 2.5976 | 2.9669 | 6.5277 | 5.00% | 45.00% | 10.00% | 58.33% | 29.6 |
| 22 | tiny-aya-water-bf16.gguf | CohereLabs/tiny-aya-water-GGUF | 1.5777 | 1.4113 | 1.4445 | 9.9833 | 15.00% | 48.33% | 5.00% | 50.00% | 29.6 |
| 23 | tiny-aya-water-f16.gguf | CohereLabs/tiny-aya-water-GGUF | 6.2966 | 6.2888 | 1.3314 | 2.7705 | 15.00% | 48.33% | 5.00% | 50.00% | 29.6 |
| 24 | LFM2-24B-A2B.Q3_K_M.gguf | mradermacher/LFM2-24B-A2B-GGUF | 6.3364 | 6.8649 | 2.5672 | 3.9174 | 10.00% | 56.67% | 20.00% | 31.67% | 29.6 |
| 25 | Ling-Coder-lite.i1-IQ3_M.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 13.5421 | 15.3022 | 0.6644 | 1.4171 | 5.00% | 46.67% | 20.00% | 46.67% | 29.6 |
| 26 | Ling-Coder-lite.i1-IQ4_NL.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 24.3246 | 24.2965 | 0.6518 | 1.1158 | 0.00% | 48.33% | 25.00% | 43.33% | 29.2 |
| 27 | qwen2.5-coder-7b-instruct-q6_k.gguf | Qwen/Qwen2.5-Coder-7B-Instruct-GGUF | 5.8176 | 6.1483 | 1.1956 | 2.9661 | 5.00% | 46.67% | 10.00% | 55.00% | 29.2 |
| 28 | Qwen3-14B-GPT-5.2-High-Reasoning-Distill.q3_k_s.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 2.3530 | 2.5285 | 1.6041 | 4.0624 | 25.00% | 18.33% | 15.00% | 58.33% | 29.2 |
| 29 | qwen2.5-coder-7b-instruct-q8_0.gguf | Qwen/Qwen2.5-Coder-7B-Instruct-GGUF | 7.2523 | 7.2361 | 0.6107 | 2.1867 | 5.00% | 45.00% | 10.00% | 56.67% | 29.2 |
| 30 | Devstral-Small-2-24B-Instruct-2512-Q3_K_L.gguf | lmstudio-community/Devstral-Small-2-24B-Instruct-2512-GGUF | 2.2775 | 2.8302 | 2.5264 | 6.9826 | 0.00% | 21.67% | 20.00% | 71.67% | 28.3 |
| 31 | Huihui-Ling-mini-2.0-abliterated.i1-Q3_K_L.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 44.7472 | 47.3020 | 0.4341 | 0.6977 | 10.00% | 41.67% | 0.00% | 61.67% | 28.3 |
| 32 | Ling-Coder-lite.i1-Q4_K_M.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 13.5022 | 15.1175 | 1.8941 | 2.7354 | 0.00% | 53.33% | 10.00% | 48.33% | 27.9 |
| 33 | LFM2.5-1.2B-Instruct-BF16.gguf | LiquidAI/LFM2.5-1.2B-Instruct-GGUF | 2.1422 | 2.2859 | 4.0274 | 7.0023 | 25.00% | 38.33% | 10.00% | 36.67% | 27.5 |
| 34 | LFM2.5-1.2B-Instruct-F16.gguf | LiquidAI/LFM2.5-1.2B-Instruct-GGUF | 14.7162 | 15.4683 | 0.4737 | 0.9506 | 25.00% | 38.33% | 10.00% | 36.67% | 27.5 |
| 35 | LFM2-24B-A2B.Q2_K.gguf | mradermacher/LFM2-24B-A2B-GGUF | 11.3713 | 13.6591 | 0.5866 | 0.9853 | 5.00% | 56.67% | 20.00% | 28.33% | 27.5 |
| 36 | Ling-Coder-lite.i1-Q5_K_S.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 17.0909 | 21.5938 | 1.7461 | 2.4249 | 5.00% | 50.00% | 15.00% | 38.33% | 27.1 |
| 37 | inclusionAI_Ling-mini-2.0-Q4_K_S.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 40.5684 | 40.4369 | 0.8203 | 1.2584 | 5.00% | 40.00% | 5.00% | 58.33% | 27.1 |
| 38 | Ling-Coder-lite.i1-Q3_K_M.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 20.0172 | 19.0624 | 0.4121 | 0.9332 | 5.00% | 45.00% | 10.00% | 46.67% | 26.7 |
| 39 | Nemotron-Cascade-14B-Thinking-Claude-4.5-Opus-Distill.iq4_nl.gguf | TeichAI/Nemotron-Cascade-14B-Thinking-Claude-4.5-Opus-High-Reasoning-Distill-GGUF | 6.7789 | 6.2501 | 0.7633 | 2.3191 | 25.00% | 11.67% | 20.00% | 50.00% | 26.7 |
| 40 | Ling-mini-2.0-Q6_K.gguf | inclusionAI/Ling-mini-2.0-GGUF | 23.9158 | 13.6525 | 1.6018 | 2.4087 | 5.00% | 45.00% | 10.00% | 45.00% | 26.2 |
| 41 | Ling-mini-2.0.Q6_K.gguf | mradermacher/Ling-mini-2.0-GGUF | 23.3610 | 24.1501 | 1.6234 | 2.6171 | 5.00% | 45.00% | 10.00% | 45.00% | 26.2 |
| 42 | Qwen3.5-9B-Q3_K_S.gguf | unsloth/Qwen3.5-9B-GGUF | 1.8171 | 2.0047 | 2.4217 | 4.9537 | 0.00% | 15.00% | 25.00% | 65.00% | 26.2 |
| 43 | inclusionAI_Ling-mini-2.0-Q3_K_L.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 46.1671 | 46.3372 | 0.3096 | 0.5471 | 15.00% | 30.00% | 0.00% | 60.00% | 26.2 |
| 44 | Ling-Coder-lite.i1-Q5_K_M.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 15.6487 | 11.6026 | 5.2806 | 6.4304 | 5.00% | 50.00% | 10.00% | 38.33% | 25.8 |
| 45 | inclusionAI_Ling-mini-2.0-IQ4_NL.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 51.2410 | 48.3286 | 0.2591 | 0.5334 | 5.00% | 50.00% | 5.00% | 43.33% | 25.8 |
| 46 | inclusionAI_Ling-mini-2.0-Q3_K_XL.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 44.7531 | 46.0997 | 0.3355 | 0.5719 | 10.00% | 31.67% | 0.00% | 61.67% | 25.8 |
| 47 | inclusionAI_Ling-mini-2.0-Q6_K_L.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 19.4811 | 28.0263 | 1.4430 | 2.5334 | 10.00% | 41.67% | 5.00% | 46.67% | 25.8 |
| 48 | inclusionAI_Ling-mini-2.0-Q4_1.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 39.6200 | 46.6260 | 0.4926 | 0.7694 | 10.00% | 36.67% | 0.00% | 56.67% | 25.8 |
| 49 | Huihui-Ling-mini-2.0-abliterated.i1-Q4_1.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 42.5288 | 46.9616 | 0.4958 | 0.7535 | 10.00% | 41.67% | 0.00% | 50.00% | 25.4 |
| 50 | Ling-Coder-lite.i1-IQ3_S.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 15.0901 | 15.0669 | 0.6170 | 1.3791 | 5.00% | 41.67% | 10.00% | 45.00% | 25.4 |
| 51 | Nemotron-Cascade-8B-Thinking-Claude-4.5-Opus-Distill.q8_0.gguf | TeichAI/Nemotron-Cascade-8B-Thinking-Claude-4.5-Opus-High-Reasoning-Distill-GGUF | 7.0487 | 7.3191 | 0.3945 | 2.0857 | 15.00% | 15.00% | 10.00% | 61.67% | 25.4 |
| 52 | Nemotron-Cascade-8B-Thinking.Q8_0.gguf | TeichAI/Nemotron-Cascade-8B-Thinking-Claude-4.5-Opus-High-Reasoning-Distill-GGUF | 6.6190 | 7.1090 | 0.3998 | 2.0898 | 15.00% | 15.00% | 10.00% | 61.67% | 25.4 |
| 53 | Qwen_Qwen3.5-9B-Q4_0.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 4.0363 | 4.0932 | 0.5057 | 2.2298 | 0.00% | 10.00% | 30.00% | 61.67% | 25.4 |
| 54 | LFM2-24B-A2B.Q3_K_S.gguf | mradermacher/LFM2-24B-A2B-GGUF | 8.4055 | 11.3072 | 1.1319 | 1.8700 | 10.00% | 35.00% | 25.00% | 31.67% | 25.4 |
| 55 | Qwen3-14B-GPT-5.2-High-Reasoning-Distill.iq4_xs.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 2.9884 | 2.5449 | 1.3722 | 2.6050 | 10.00% | 13.33% | 15.00% | 61.67% | 25.0 |
| 56 | Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q4_K_M.gguf | HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive | 2.8328 | 2.8425 | 1.0612 | 3.1771 | 0.00% | 10.00% | 25.00% | 65.00% | 25.0 |
| 57 | GPT-OSS-Swallow-20B-RL-v0.1-Q3_K_L.gguf | mmnga-o/GPT-OSS-Swallow-20B-RL-v0.1-gguf | 8.8377 | 8.9720 | 4.3291 | 4.9654 | 30.00% | 15.00% | 5.00% | 48.33% | 24.6 |
| 58 | Huihui-Ling-mini-2.0-abliterated.i1-Q3_K_M.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 46.7636 | 50.0205 | 0.3043 | 0.5418 | 0.00% | 40.00% | 0.00% | 58.33% | 24.6 |
| 59 | Ling-Coder-lite.i1-Q3_K_S.gguf | mradermacher/Ling-Coder-lite-i1-GGUF | 24.9258 | 26.0106 | 0.4194 | 0.8889 | 5.00% | 51.67% | 15.00% | 26.67% | 24.6 |
| 60 | Ling-mini-2.0.IQ4_XS.gguf | mradermacher/Ling-mini-2.0-GGUF | 47.4974 | 49.6456 | 0.3617 | 0.5773 | 0.00% | 51.67% | 0.00% | 46.67% | 24.6 |
| 61 | Qwen3.5-9B-Q4_0.gguf | unsloth/Qwen3.5-9B-GGUF | 3.9738 | 3.9831 | 0.5685 | 2.2648 | 0.00% | 10.00% | 30.00% | 58.33% | 24.6 |
| 62 | inclusionAI_Ling-mini-2.0-IQ4_XS.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 48.7321 | 49.9764 | 0.4314 | 0.6673 | 5.00% | 50.00% | 5.00% | 38.33% | 24.6 |
| 63 | inclusionAI_Ling-mini-2.0-Q4_K_M.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 41.1503 | 41.9180 | 0.4106 | 0.7620 | 10.00% | 50.00% | 0.00% | 38.33% | 24.6 |
| 64 | Huihui-Ling-mini-2.0-abliterated.i1-Q4_K_S.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 49.7664 | 50.9208 | 0.3825 | 0.6799 | 5.00% | 41.67% | 0.00% | 50.00% | 24.2 |
| 65 | Qwen3-14B-GPT-5.2-Distill.iq4_nl.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 4.0868 | 4.0773 | 0.9961 | 2.2783 | 15.00% | 13.33% | 15.00% | 53.33% | 24.2 |
| 66 | Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf | HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive | 2.0596 | 2.0388 | 1.6798 | 4.3698 | 0.00% | 13.33% | 30.00% | 53.33% | 24.2 |
| 67 | inclusionAI_Ling-mini-2.0-Q5_K_L.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 34.1054 | 28.4821 | 1.1611 | 1.6150 | 5.00% | 46.67% | 5.00% | 40.00% | 24.2 |
| 68 | Huihui-Ling-mini-2.0-abliterated.i1-IQ4_XS.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 43.1980 | 45.4173 | 0.3574 | 0.5717 | 5.00% | 48.33% | 0.00% | 41.67% | 23.8 |
| 69 | Huihui-Ling-mini-2.0-abliterated.i1-Q5_K_M.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 37.0431 | 25.9076 | 1.3095 | 1.7583 | 5.00% | 41.67% | 0.00% | 48.33% | 23.8 |
| 70 | Qwen3-14B-GPT-5.2-Distill.q4_k_m.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 3.7102 | 3.6092 | 1.1571 | 2.5365 | 20.00% | 15.00% | 15.00% | 45.00% | 23.8 |
| 71 | GPT-OSS-Swallow-20B-RL-v0.1-Q4_0.gguf | mmnga-o/GPT-OSS-Swallow-20B-RL-v0.1-gguf | 9.4349 | 11.0214 | 3.0715 | 3.7612 | 25.00% | 16.67% | 10.00% | 41.67% | 23.3 |
| 72 | Huihui-Ling-mini-2.0-abliterated.i1-IQ3_XXS.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 32.1211 | 32.3675 | 0.3370 | 0.6693 | 0.00% | 35.00% | 10.00% | 48.33% | 23.3 |
| 73 | Huihui-Ling-mini-2.0-abliterated.i1-Q5_K_S.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 45.0309 | 45.9212 | 0.4834 | 0.7313 | 5.00% | 41.67% | 0.00% | 46.67% | 23.3 |
| 74 | gpt-oss-20b-glm-4.6-distill-bf16.gguf | TeichAI/gpt-oss-20b-glm-4.6-distill-GGUF | 1.0688 | 1.0931 | 4.6648 | 7.4988 | 25.00% | 3.33% | 15.00% | 50.00% | 23.3 |
| 75 | gpt-oss-20b-glm-4.6-distill-f16.gguf | TeichAI/gpt-oss-20b-glm-4.6-distill-GGUF | 3.8221 | 3.9594 | 4.1389 | 4.8853 | 25.00% | 3.33% | 15.00% | 50.00% | 23.3 |
| 76 | mistralai_Devstral-Small-2-24B-Instruct-2512-Q3_K_L.gguf | bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF | 2.5200 | 2.3172 | 2.4080 | 6.6253 | 0.00% | 6.67% | 20.00% | 66.67% | 23.3 |
| 77 | Ling-mini-2.0.Q5_K_S.gguf | mradermacher/Ling-mini-2.0-GGUF | 41.8266 | 44.1412 | 0.5435 | 0.7920 | 5.00% | 43.33% | 0.00% | 43.33% | 22.9 |
| 78 | inclusionAI_Ling-mini-2.0-Q4_K_L.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 37.5287 | 39.1054 | 0.4862 | 0.8609 | 5.00% | 43.33% | 5.00% | 38.33% | 22.9 |
| 79 | nanbeige4.1-python-deepthink-fp16.gguf | deltakitsune/Nanbeige-4.1-Python-DeepThink-3B | 6.7471 | 6.7378 | 0.7510 | 2.0288 | 10.00% | 20.00% | 10.00% | 51.67% | 22.9 |
| 80 | Huihui-Ling-mini-2.0-abliterated.i1-Q4_K_M.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 36.4292 | 37.4090 | 0.4137 | 0.8073 | 5.00% | 46.67% | 0.00% | 40.00% | 22.9 |
| 81 | gpt-oss-20b-glm-4.6-distill-q8_0.gguf | TeichAI/gpt-oss-20b-glm-4.6-distill-GGUF | 9.0254 | 9.4644 | 3.1292 | 3.6791 | 20.00% | 3.33% | 15.00% | 53.33% | 22.9 |
| 82 | mistralai_Devstral-Small-2-24B-Instruct-2512-Q4_0.gguf | bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF | 3.4268 | 3.3014 | 2.6874 | 5.5445 | 0.00% | 33.33% | 10.00% | 48.33% | 22.9 |
| 83 | Huihui-Ling-mini-2.0-abliterated.i1-IQ4_NL.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 44.7265 | 43.9510 | 0.2593 | 0.5337 | 0.00% | 45.00% | 5.00% | 40.00% | 22.5 |
| 84 | Qwen3.5-9B-IQ4_XS.gguf | unsloth/Qwen3.5-9B-GGUF | 2.8629 | NA | 1.4357 | 3.1437 | 0.00% | 10.00% | 25.00% | 55.00% | 22.5 |
| 85 | Qwen3.5-9B-Q3_K_M.gguf | unsloth/Qwen3.5-9B-GGUF | 2.4079 | 2.3508 | 1.6312 | 3.7377 | 0.00% | 13.33% | 25.00% | 51.67% | 22.5 |
| 86 | inclusionAI_Ling-mini-2.0-Q5_K_M.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 37.6751 | 37.6650 | 1.3262 | 1.7806 | 5.00% | 46.67% | 0.00% | 38.33% | 22.5 |
| 87 | Nemotron-Cascade-14B-Thinking-Claude-4.5-Opus-Distill.q4_k_m.gguf | TeichAI/Nemotron-Cascade-14B-Thinking-Claude-4.5-Opus-High-Reasoning-Distill-GGUF | 4.6044 | 4.1704 | 1.2234 | 3.4274 | 15.00% | 10.00% | 10.00% | 53.33% | 22.1 |
| 88 | Qwen3-14B-GPT-5.2-High-Reasoning-Distill.iq4_nl.gguf | TeichAI/Qwen3-14B-GPT-5.2-High-Reasoning-Distill-GGUF | 2.7653 | 2.7633 | 0.9604 | 2.2355 | 5.00% | 20.00% | 5.00% | 58.33% | 22.1 |
| 89 | inclusionAI_Ling-mini-2.0-Q5_K_S.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 38.5676 | 45.0414 | 0.4959 | 0.7647 | 5.00% | 46.67% | 0.00% | 36.67% | 22.1 |
| 90 | Ling-mini-2.0.Q4_K_S.gguf | mradermacher/Ling-mini-2.0-GGUF | 41.4219 | 41.9427 | 0.4141 | 0.7309 | 5.00% | 43.33% | 0.00% | 38.33% | 21.7 |
| 91 | Ling-mini-2.0.Q5_K_M.gguf | mradermacher/Ling-mini-2.0-GGUF | 39.4459 | 40.3414 | 0.6775 | 0.9837 | 5.00% | 43.33% | 0.00% | 38.33% | 21.7 |
| 92 | Nemotron-Cascade-8B-Thinking-Claude-4.5-Opus-Distill.q4_k_m.gguf | TeichAI/Nemotron-Cascade-8B-Thinking-Claude-4.5-Opus-High-Reasoning-Distill-GGUF | 8.4103 | 8.7868 | 0.4660 | 1.8573 | 5.00% | 13.33% | 10.00% | 58.33% | 21.7 |
| 93 | Qwen3.5-9B-UD-Q4_K_XL.gguf | unsloth/Qwen3.5-9B-GGUF | 2.6871 | 2.6339 | 1.2633 | 3.3493 | 0.00% | 11.67% | 20.00% | 55.00% | 21.7 |
| 94 | inclusionAI_Ling-mini-2.0-Q6_K.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 29.2826 | 14.8480 | 1.6699 | 2.4159 | 5.00% | 41.67% | 0.00% | 40.00% | 21.7 |
| 95 | Huihui-Ling-mini-2.0-abliterated.i1-IQ3_XS.gguf | mradermacher/Huihui-Ling-mini-2.0-abliterated-i1-GGUF | 29.2746 | 27.0336 | 0.2833 | 0.6621 | 0.00% | 40.00% | 0.00% | 46.67% | 21.7 |
| 96 | Qwen3.5-9B-Q5_K_M.gguf | unsloth/Qwen3.5-9B-GGUF | 2.1796 | 2.1586 | 1.8871 | 4.1292 | 0.00% | 11.67% | 35.00% | 40.00% | 21.7 |
| 97 | Qwen3.5-9B-UD-IQ3_XXS.gguf | unsloth/Qwen3.5-9B-GGUF | 1.4893 | 1.4596 | 3.2478 | 6.0431 | 0.00% | 10.00% | 20.00% | 56.67% | 21.7 |
| 98 | gpt-oss-9b-q8_0.gguf | squ11z1/gpt-oss-nano | 7.9722 | 8.4680 | 1.2882 | 2.0016 | 20.00% | 8.33% | 10.00% | 46.67% | 21.2 |
| 99 | Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5.Q8_0.gguf | crownelius/Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5 | 6.9403 | 6.9356 | 0.4816 | 2.0172 | 0.00% | 0.00% | 30.00% | 53.33% | 20.8 |
| 100 | Qwen_Qwen3.5-9B-Q4_K_S.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 3.1749 | 3.1091 | 0.8445 | 2.8347 | 0.00% | 11.67% | 15.00% | 55.00% | 20.4 |
| 101 | Qwen3.5-9B-IQ4_NL.gguf | unsloth/Qwen3.5-9B-GGUF | 3.8342 | 3.7981 | 0.6561 | 2.3473 | 0.00% | 10.00% | 15.00% | 56.67% | 20.4 |
| 102 | Qwen3.5-9B-Q4_1.gguf | unsloth/Qwen3.5-9B-GGUF | 2.5915 | 2.5522 | 1.6290 | 3.4729 | 0.00% | 11.67% | 35.00% | 35.00% | 20.4 |
| 103 | Qwen_Qwen3.5-9B-IQ4_XS.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.8684 | 2.8471 | 1.3927 | 3.1377 | 0.00% | 11.67% | 20.00% | 50.00% | 20.4 |
| 104 | inclusionAI_Ling-mini-2.0-Q4_0.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 47.8941 | 37.9757 | 0.7811 | 1.1748 | 10.00% | 36.67% | 0.00% | 33.33% | 20.0 |
| 105 | Qwen3.5-9B.Q4_K_H.gguf | steampunque/Qwen3.5-9B-MP-GGUF | 2.8178 | 2.7556 | 1.1202 | 3.1940 | 0.00% | 10.00% | 30.00% | 38.33% | 19.6 |
| 106 | phi-4-reasoning-vision-q4_k_s.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | 4.8125 | 4.7690 | 0.9961 | 3.2543 | 0.00% | 58.33% | 0.00% | 20.00% | 19.6 |
| 107 | Qwen_Qwen3.5-9B-IQ4_NL.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 3.8881 | 3.8777 | 0.5943 | 2.3147 | 0.00% | 8.33% | 10.00% | 60.00% | 19.6 |
| 108 | Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5.Q5_K_M.gguf | crownelius/Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5 | 5.6107 | 5.5854 | 1.0373 | 2.4952 | 0.00% | 0.00% | 25.00% | 51.67% | 19.2 |
| 109 | Qwen3.5-9B-Q6_K.gguf | lmstudio-community/Qwen3.5-9B-GGUF | 2.0969 | 2.0621 | 1.6491 | 4.2921 | 0.00% | 8.33% | 15.00% | 53.33% | 19.2 |
| 110 | Qwen3.5-9B-Q4_K_S.gguf | unsloth/Qwen3.5-9B-GGUF | 3.0669 | 3.0640 | 0.9148 | 2.9346 | 0.00% | 11.67% | 20.00% | 45.00% | 19.2 |
| 111 | Qwen_Qwen3.5-9B-Q4_K_L.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.8480 | 2.8511 | 1.0608 | 3.1601 | 0.00% | 10.00% | 10.00% | 56.67% | 19.2 |
| 112 | inclusionAI_Ling-mini-2.0-IQ3_M.gguf | bartowski/inclusionAI_Ling-mini-2.0-GGUF | 31.7144 | 32.3336 | 0.2973 | 0.6079 | 5.00% | 38.33% | 0.00% | 31.67% | 18.8 |
| 113 | \*mistralai_Devstral-Small-2-24B-Instruct-2512-Q4_1.gguf | bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF | 0.0542 | 0.0536 | 17.4830 | 267.3234 | 0.00% | 0.00% | 15.00% | 60.00% | 18.8 |
| 114 | \*Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5.BF16.gguf | crownelius/Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5 | 0.5318 | 0.5318 | 15.0294 | 26.3262 | 0.00% | 0.00% | 20.00% | 53.33% | 18.3 |
| 115 | Ling-mini-2.0-Q4_K_M.gguf | inclusionAI/Ling-mini-2.0-GGUF | 39.9336 | 46.1113 | 0.4382 | 0.8347 | 0.00% | 38.33% | 0.00% | 35.00% | 18.3 |
| 116 | Ling-mini-2.0.Q4_K_M.gguf | mradermacher/Ling-mini-2.0-GGUF | 36.7990 | 44.4606 | 0.3824 | 0.7542 | 0.00% | 38.33% | 0.00% | 35.00% | 18.3 |
| 117 | Qwen_Qwen3.5-9B-Q6_K.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.1860 | 2.1839 | 1.5031 | 4.1172 | 0.00% | 11.67% | 20.00% | 41.67% | 18.3 |
| 118 | Qwen_Qwen3.5-9B-Q6_K_L.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.2456 | 2.2181 | 1.4654 | 4.0079 | 0.00% | 10.00% | 20.00% | 43.33% | 18.3 |
| 119 | Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q8_0.gguf | HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive | 2.7111 | 2.7918 | 0.9588 | 3.3251 | 0.00% | 11.67% | 25.00% | 36.67% | 18.3 |
| 120 | Qwen_Qwen3.5-9B-Q4_1.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.5744 | 2.6086 | 1.6048 | 3.4960 | 0.00% | 10.00% | 20.00% | 41.67% | 17.9 |
| 121 | Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5.Q4_K_M.gguf | crownelius/Crow-4B-Opus-4.6-Distill-Heretic_Qwen3.5 | 6.8873 | 6.8824 | 0.6215 | 2.0327 | 0.00% | 1.67% | 15.00% | 51.67% | 17.1 |
| 122 | Qwen3.5-9B-UD-Q3_K_XL.gguf | unsloth/Qwen3.5-9B-GGUF | 2.0038 | 2.0685 | 1.9399 | 4.4919 | 0.00% | 11.67% | 15.00% | 41.67% | 17.1 |
| 123 | Qwen_Qwen3.5-9B-Q5_K_M.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.2387 | 2.2292 | 1.8026 | 4.0202 | 0.00% | 10.00% | 20.00% | 38.33% | 17.1 |
| 124 | Qwen3.5-2B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-2B-GGUF | 10.6239 | 10.6096 | 0.2468 | 0.8472 | 0.00% | 13.33% | 15.00% | 38.33% | 16.7 |
| 125 | Qwen3.5-9B-Q8_0.gguf | lmstudio-community/Qwen3.5-9B-GGUF | 2.6731 | 2.7878 | 1.0222 | 3.3803 | 0.00% | 10.00% | 15.00% | 41.67% | 16.7 |
| 126 | Qwen_Qwen3.5-9B-Q8_0.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.6301 | 2.6284 | 0.8903 | 3.4254 | 0.00% | 10.00% | 15.00% | 41.67% | 16.7 |
| 127 | phi-4-reasoning-vision-q5_k_s.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | 4.6354 | 4.3270 | 1.9206 | 4.2545 | 0.00% | 53.33% | 0.00% | 13.33% | 16.7 |
| 128 | Qwen3.5-9B-Q5_K_S.gguf | unsloth/Qwen3.5-9B-GGUF | 2.2846 | 2.2368 | 1.8912 | 3.9395 | 0.00% | 8.33% | 25.00% | 33.33% | 16.7 |
| 129 | Qwen3.5-9B-UD-Q8_K_XL.gguf | unsloth/Qwen3.5-9B-GGUF | 1.9490 | 1.8904 | 1.7520 | 4.6273 | 0.00% | 11.67% | 15.00% | 40.00% | 16.7 |
| 130 | Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill.bf16.gguf | TeichAI/Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill-GGUF | 1.1763 | 1.1307 | 7.9154 | 18.1766 | 5.00% | 5.00% | 10.00% | 43.33% | 15.8 |
| 131 | Qwen3.5-2B.i1-Q4_K_M.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 10.5134 | 10.5139 | 0.2489 | 0.8561 | 0.00% | 13.33% | 5.00% | 45.00% | 15.8 |
| 132 | Qwen3.5-9B-UD-Q5_K_XL.gguf | unsloth/Qwen3.5-9B-GGUF | 2.3145 | 2.2724 | 1.6479 | 3.8886 | 0.00% | 11.67% | 10.00% | 41.67% | 15.8 |
| 133 | Qwen3.5-2B-Q4_K_S.gguf | unsloth/Qwen3.5-2B-GGUF | 10.7885 | 10.8671 | 0.2201 | 0.8342 | 0.00% | 11.67% | 15.00% | 36.67% | 15.8 |
| 134 | Qwen_Qwen3.5-9B-Q5_K_L.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.2459 | 2.2279 | 1.8170 | 4.0073 | 0.00% | 8.33% | 20.00% | 35.00% | 15.8 |
| 135 | ggml-model-Q4_K_M.gguf |  | 3.8180 | 3.7491 | 7.8247 | 10.4099 | 15.00% | 8.33% | 0.00% | 40.00% | 15.8 |
| 136 | Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill.q8_0.gguf | TeichAI/Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill-GGUF | 15.7523 | 15.3764 | 0.2203 | 1.1286 | 5.00% | 5.00% | 10.00% | 41.67% | 15.4 |
| 137 | Qwen3.5-2B.i1-Q4_K_S.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 11.2415 | 11.2306 | 0.2030 | 0.8006 | 0.00% | 13.33% | 5.00% | 43.33% | 15.4 |
| 138 | Qwen3.5-4B-heretic-imatrix-Q4_0.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 6.0418 | 6.0374 | 0.3063 | 1.4896 | 0.00% | 13.33% | 5.00% | 43.33% | 15.4 |
| 139 | phi-4-reasoning-vision-q4_K_M.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | 4.7340 | 4.3983 | 1.2082 | 3.6313 | 0.00% | 51.67% | 0.00% | 10.00% | 15.4 |
| 140 | phi-4-reasoning-vision-q5_k_m.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | 3.6007 | 3.8254 | 1.9390 | 4.5986 | 0.00% | 48.33% | 0.00% | 13.33% | 15.4 |
| 141 | qwen3.5-9b-null-space-abliterated-q6_k.gguf | jwest33/qwen3.5-9b-null-space-abliterated-GGUF | 2.0600 | 2.0425 | 1.6806 | 4.3689 | 0.00% | 13.33% | 20.00% | 28.33% | 15.4 |
| 142 | Qwen3.5-2B.i1-IQ4_NL.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 13.4602 | 13.4427 | 0.1455 | 0.6686 | 0.00% | 13.33% | 15.00% | 33.33% | 15.4 |
| 143 | Qwen3.5-2B-Q4_0.gguf | unsloth/Qwen3.5-2B-GGUF | 13.3080 | 13.2891 | 0.1437 | 0.6763 | 0.00% | 13.33% | 15.00% | 31.67% | 15.0 |
| 144 | Qwen3.5-9B-UD-Q6_K_XL.gguf | unsloth/Qwen3.5-9B-GGUF | 2.1112 | 2.1284 | 1.6298 | 4.2629 | 0.00% | 8.33% | 10.00% | 41.67% | 15.0 |
| 145 | Qwen_Qwen3.5-9B-Q5_K_S.gguf | bartowski/Qwen_Qwen3.5-9B-GGUF | 2.2775 | 2.2535 | 1.8916 | 3.9518 | 0.00% | 11.67% | 15.00% | 33.33% | 15.0 |
| 146 | Qwen3.5-2B-IQ4_NL.gguf | unsloth/Qwen3.5-2B-GGUF | 13.0621 | 12.8939 | 0.1644 | 0.6890 | 0.00% | 13.33% | 10.00% | 35.00% | 14.6 |
| 147 | qwen3-4b-thinking-2507.F16.gguf | TeichAI/Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill-GGUF | 7.5548 | 7.3958 | 0.9576 | 2.6511 | 5.00% | 5.00% | 5.00% | 43.33% | 14.6 |
| 148 | Ling-mini-2.0-Q2_K.gguf | inclusionAI/Ling-mini-2.0-GGUF | 42.8914 | 41.5145 | 0.2690 | 0.4860 | 0.00% | 30.00% | 0.00% | 28.33% | 14.6 |
| 149 | Qwen3.5-9B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-9B-GGUF | 2.8146 | 2.8131 | 1.0669 | 3.1977 | 0.00% | 8.33% | 15.00% | 35.00% | 14.6 |
| 150 | GLM-4.7-Flash-Uncen-Hrt-NEO-CODE-MAX-imat-D_AU-IQ3_M.gguf | DavidAU/GLM-4.7-Flash-Uncensored-Heretic-NEO-CODE-Imatrix-MAX-GGUF | 2.2089 | 2.1602 | 1.8866 | 4.1533 | 0.00% | 0.00% | 20.00% | 36.67% | 14.2 |
| 151 | Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill.q4_k_m.gguf | TeichAI/Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill-GGUF | 17.4602 | 16.7520 | 0.2468 | 1.0326 | 0.00% | 10.00% | 0.00% | 46.67% | 14.2 |
| 152 | Qwen3.5-2B.i1-Q4_0.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 13.8116 | 13.8364 | 0.1282 | 0.6516 | 0.00% | 11.67% | 10.00% | 35.00% | 14.2 |
| 153 | Qwen3-8B-Q6_K.gguf | Qwen/Qwen3-8B-GGUF | 5.3310 | 4.9952 | 1.3109 | 2.7192 | 0.00% | 3.33% | 0.00% | 51.67% | 13.8 |
| 154 | Qwen3.5-2B-Q5_K_S.gguf | unsloth/Qwen3.5-2B-GGUF | 8.6313 | 8.6581 | 0.4245 | 1.0427 | 0.00% | 10.00% | 10.00% | 35.00% | 13.8 |
| 155 | Qwen3.5-35B-A3B-UD-IQ2_M.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 3.3513 | 3.4083 | 1.3374 | 2.8379 | 0.00% | 10.00% | 45.00% | 0.00% | 13.8 |
| 156 | Qwen_Qwen3.5-35B-A3B-IQ3_XS.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 2.0368 | 2.0524 | 3.8620 | 6.2538 | 5.00% | 10.00% | 40.00% | 0.00% | 13.8 |
| 157 | Qwen3-8B-Q8_0.gguf | Qwen/Qwen3-8B-GGUF | 6.3467 | 6.2934 | 0.8483 | 2.0884 | 0.00% | 3.33% | 0.00% | 50.00% | 13.3 |
| 158 | \*Qwen3.5-27B-UD-IQ3_XXS.gguf | unsloth/Qwen3.5-27B-GGUF | 0.4213 | 0.3872 | 12.5484 | 21.4678 | 0.00% | 13.33% | 40.00% | 0.00% | 13.3 |
| 159 | Qwen_Qwen3.5-35B-A3B-IQ2_M.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.3018 | 3.6545 | 1.2444 | 2.7662 | 0.00% | 11.67% | 40.00% | 1.67% | 13.3 |
| 160 | Qwen3.5-2B-Q4_1.gguf | unsloth/Qwen3.5-2B-GGUF | 9.6449 | 9.7576 | 0.3559 | 0.9333 | 0.00% | 15.00% | 5.00% | 31.67% | 12.9 |
| 161 | Qwen3.5-2B-Q6_K.gguf | lmstudio-community/Qwen3.5-2B-GGUF | 8.6728 | 8.6108 | 0.3424 | 1.0377 | 0.00% | 11.67% | 5.00% | 35.00% | 12.9 |
| 162 | Qwen3.5-2B-Q8_0.gguf | lmstudio-community/Qwen3.5-2B-GGUF | 10.8192 | 10.7967 | 0.1839 | 0.8319 | 0.00% | 11.67% | 5.00% | 35.00% | 12.9 |
| 163 | phi-4-reasoning-vision-q6_k.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | 2.8199 | 2.9401 | 2.1462 | 5.4721 | 0.00% | 45.00% | 0.00% | 6.67% | 12.9 |
| 164 | Qwen3.5-27B-Q3_K_S.gguf | unsloth/Qwen3.5-27B-GGUF | 0.6118 | 0.5382 | 7.7920 | 14.7111 | 0.00% | 0.00% | 50.00% | 0.00% | 12.5 |
| 165 | Qwen3.5-35B-A3B-UD-Q2_K_XL.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 3.3811 | 3.2458 | 1.5488 | 3.0509 | 0.00% | 10.00% | 40.00% | 0.00% | 12.5 |
| 166 | ggml-model-Q3_K_M.gguf |  | 9.3976 | 9.8715 | 4.0421 | 4.6873 | 10.00% | 11.67% | 0.00% | 28.33% | 12.5 |
| 167 | Qwen3.5-2B-Q5_K_M.gguf | unsloth/Qwen3.5-2B-GGUF | 8.5584 | 8.5323 | 0.4041 | 1.0516 | 0.00% | 10.00% | 0.00% | 38.33% | 12.1 |
| 168 | Qwen3.5-35B-A3B-UD-IQ2_XXS.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 3.2888 | 3.7104 | 1.4168 | 2.9327 | 0.00% | 13.33% | 35.00% | 0.00% | 12.1 |
| 169 | Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ2_M.gguf | HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive | 3.2624 | 3.1706 | 1.3138 | 2.8025 | 0.00% | 13.33% | 35.00% | 0.00% | 12.1 |
| 170 | Qwen_Qwen3.5-35B-A3B-Q2_K_L.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.7061 | 3.8353 | 1.2755 | 2.6581 | 0.00% | 8.33% | 40.00% | 0.00% | 12.1 |
| 171 | glm-4.7-flash-claude-4.5-opus.iq3_xs.gguf | TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill-GGUF | 9.1172 | 8.9179 | 0.8661 | 2.1619 | 5.00% | 11.67% | 0.00% | 31.67% | 12.1 |
| 172 | qwen3.5-9b-null-space-abliterated-q8_0.gguf | jwest33/qwen3.5-9b-null-space-abliterated-GGUF | 2.6512 | 2.7994 | 0.9372 | 3.4023 | 0.00% | 11.67% | 5.00% | 31.67% | 12.1 |
| 173 | Qwen3.5-35B-A3B-UD-Q3_K_XL.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 2.3298 | 2.3226 | 3.6484 | 6.0833 | 0.00% | 11.67% | 35.00% | 0.00% | 11.7 |
| 174 | Qwen_Qwen3.5-35B-A3B-Q2_K.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.8560 | 4.1703 | 1.0980 | 2.4616 | 0.00% | 8.33% | 35.00% | 3.33% | 11.7 |
| 175 | Qwen_Qwen3.5-35B-A3B-Q3_K_S.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 2.6052 | 2.5577 | 2.6178 | 4.6266 | 0.00% | 10.00% | 35.00% | 0.00% | 11.2 |
| 176 | Qwen_Qwen3.5-35B-A3B-IQ3_XXS.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 2.4333 | 2.4020 | 2.1453 | 3.9737 | 5.00% | 8.33% | 30.00% | 0.00% | 10.8 |
| 177 | Qwen3.5-27B-UD-IQ2_XXS.gguf | unsloth/Qwen3.5-27B-GGUF | 3.9724 | 4.0213 | 0.7177 | 3.2733 | 0.00% | 11.67% | 30.00% | 0.00% | 10.4 |
| 178 | Qwen_Qwen3.5-27B-IQ2_M.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 2.7238 | 2.7252 | 0.6825 | 3.3042 | 0.00% | 11.67% | 30.00% | 0.00% | 10.4 |
| 179 | Qwen3.5-35B-A3B-Q2_K_L.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 3.5611 | 3.6863 | 1.3787 | 2.8116 | 0.00% | 10.00% | 30.00% | 0.00% | 10.0 |
| 180 | Qwen_Qwen3.5-27B-Q3_K_S.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.5929 | 0.5847 | 8.1024 | 15.1853 | 0.00% | 0.00% | 40.00% | 0.00% | 10.0 |
| 181 | Qwen3.5-0.8B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-0.8B-GGUF | 17.6339 | 17.6287 | 0.1269 | 0.5104 | 0.00% | 13.33% | 0.00% | 25.00% | 9.6 |
| 182 | Qwen3.5-2B.i1-Q6_K.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 8.7394 | 8.6944 | 0.3382 | 1.0298 | 0.00% | 10.00% | 5.00% | 21.67% | 9.2 |
| 183 | Qwen3.5-35B-A3B-UD-IQ3_S.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 2.4601 | 2.4713 | 2.4454 | 4.5084 | 5.00% | 11.67% | 20.00% | 0.00% | 9.2 |
| 184 | Qwen_Qwen3.5-35B-A3B-IQ2_S.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.7181 | 3.8285 | 1.0166 | 2.4291 | 0.00% | 11.67% | 25.00% | 0.00% | 9.2 |
| 185 | \*Qwen3.5-35B-A3B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-35B-A3B-GGUF | 1.1510 | 1.4402 | 13.0595 | 17.9214 | 5.00% | 0.00% | 30.00% | 0.00% | 8.8 |
| 186 | Qwen_Qwen3.5-27B-IQ2_S.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 2.7193 | 2.7681 | 0.7331 | 3.3109 | 0.00% | 15.00% | 20.00% | 0.00% | 8.8 |
| 187 | glm-4.7-flash-claude-4.5-opus.q3_k_s.gguf | TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill-GGUF | 13.1264 | 12.5699 | 0.9636 | 1.8598 | 5.00% | 11.67% | 0.00% | 18.33% | 8.8 |
| 188 | Qwen_Qwen3.5-27B-Q2_K_L.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.6942 | 0.5382 | 7.6201 | 13.3453 | 0.00% | 13.33% | 20.00% | 0.00% | 8.3 |
| 189 | Qwen_Qwen3.5-4B-Q4_1.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 4.0720 | 4.0516 | 0.9047 | 2.2102 | 0.00% | 8.33% | 25.00% | 0.00% | 8.3 |
| 190 | Qwen_Qwen3.5-35B-A3B-IQ1_M.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 9.2056 | 9.3820 | 0.2791 | 0.9781 | 0.00% | 16.67% | 15.00% | 0.00% | 7.9 |
| 191 | Qwen3.5-4B-Q4_0.gguf | unsloth/Qwen3.5-4B-GGUF | 5.8149 | 5.7623 | 0.3616 | 1.5479 | 0.00% | 10.00% | 20.00% | 0.00% | 7.5 |
| 192 | Qwen_Qwen3.5-35B-A3B-Q3_K_M.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 2.4105 | 2.3928 | 3.9147 | 6.0526 | 0.00% | 10.00% | 20.00% | 0.00% | 7.5 |
| 193 | Qwen3.5-27B-UD-Q2_K_XL.gguf | unsloth/Qwen3.5-27B-GGUF | 0.6373 | 0.6275 | 7.6373 | 14.1631 | 0.00% | 13.33% | 15.00% | 0.00% | 7.1 |
| 194 | Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ3_M.gguf | HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive | 2.0835 | 2.1398 | 3.1247 | 5.4132 | 0.00% | 13.33% | 15.00% | 0.00% | 7.1 |
| 195 | \*Qwen_Qwen3.5-27B-IQ3_XXS.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.3766 | 0.4116 | 14.1526 | 24.1141 | 0.00% | 13.33% | 15.00% | 0.00% | 7.1 |
| 196 | Qwen_Qwen3.5-27B-Q2_K.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.7801 | 0.7659 | 5.6421 | 11.5677 | 0.00% | 13.33% | 15.00% | 0.00% | 7.1 |
| 197 | Qwen3.5-4B-Q4_K_S.gguf | unsloth/Qwen3.5-4B-GGUF | 4.6822 | 4.6855 | 0.5497 | 1.9222 | 0.00% | 8.33% | 20.00% | 0.00% | 7.1 |
| 198 | Qwen_Qwen3.5-35B-A3B-IQ2_XXS.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.5993 | 3.3257 | 1.0313 | 2.5044 | 0.00% | 8.33% | 20.00% | 0.00% | 7.1 |
| 199 | Qwen3.5-35B-A3B-UD-IQ4_XS.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | NA | 2.2643 | NA | NA | 0.00% | 0.00% | 25.00% | 0.00% | 6.2 |
| 200 | Qwen_Qwen3.5-27B-IQ2_XS.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 2.7872 | 2.7705 | 0.6731 | 3.2291 | 0.00% | 15.00% | 10.00% | 0.00% | 6.2 |
| 201 | Qwen_Qwen3.5-27B-IQ2_XXS.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 2.8436 | 2.8324 | 0.6749 | 3.1650 | 0.00% | 13.33% | 10.00% | 0.00% | 5.8 |
| 202 | Qwen3.5-0.8B-Q4_0.gguf | unsloth/Qwen3.5-0.8B-GGUF | 20.2209 | 20.2357 | 0.0863 | 0.4451 | 0.00% | 15.00% | 0.00% | 6.67% | 5.4 |
| 203 | Qwen3.5-4B-heretic-imatrix-Q4_K_S.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.7949 | 4.7975 | 0.5050 | 1.8770 | 0.00% | 6.67% | 15.00% | 0.00% | 5.4 |
| 204 | Qwen3.5-0.8B-IQ4_XS.gguf | unsloth/Qwen3.5-0.8B-GGUF | 17.4546 | 17.4631 | 0.1517 | 0.5156 | 0.00% | 11.67% | 10.00% | 0.00% | 5.4 |
| 205 | Qwen3.5-0.8B-Q8_0.gguf | lmstudio-community/Qwen3.5-0.8B-GGUF | 18.0749 | 17.9288 | 0.1002 | 0.4979 | 0.00% | 15.00% | 5.00% | 0.00% | 5.0 |
| 206 | Qwen3.5-27B-Q3_K_M.gguf | unsloth/Qwen3.5-27B-GGUF | 0.6321 | 0.6221 | 8.1337 | 14.6437 | 0.00% | 0.00% | 20.00% | 0.00% | 5.0 |
| 207 | Qwen3.5-35B-A3B-UD-IQ3_XXS.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | 2.6490 | 2.5310 | 1.9092 | 3.8985 | 0.00% | 10.00% | 10.00% | 0.00% | 5.0 |
| 208 | Qwen3.5-4B-Q4_1.gguf | unsloth/Qwen3.5-4B-GGUF | 4.0398 | 4.0220 | 0.9283 | 2.2278 | 0.00% | 10.00% | 10.00% | 0.00% | 5.0 |
| 209 | Qwen3.5-4B-heretic-imatrix-Q4_0-pure.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 6.4829 | 6.5269 | 0.2926 | 1.3883 | 0.00% | 10.00% | 10.00% | 0.00% | 5.0 |
| 210 | Qwen3.5-4B-heretic-imatrix-Q4_1.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.0725 | 4.0755 | 0.9178 | 2.2100 | 0.00% | 10.00% | 10.00% | 0.00% | 5.0 |
| 211 | \*Qwen_Qwen3.5-27B-IQ3_XS.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.3472 | 0.3480 | 14.7483 | 25.9284 | 0.00% | 0.00% | 20.00% | 0.00% | 5.0 |
| 212 | Qwen_Qwen3.5-27B-Q3_K_M.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.6743 | 0.6243 | 6.8247 | 13.3750 | 0.00% | 0.00% | 20.00% | 0.00% | 5.0 |
| 213 | Qwen3.5-0.8B-IQ4_NL.gguf | unsloth/Qwen3.5-0.8B-GGUF | 19.8624 | 19.8514 | 0.0938 | 0.4531 | 0.00% | 13.33% | 5.00% | 0.00% | 4.6 |
| 214 | Qwen3.5-0.8B-Q5_K_M.gguf | unsloth/Qwen3.5-0.8B-GGUF | 15.6506 | 15.6337 | 0.1752 | 0.5751 | 0.00% | 13.33% | 5.00% | 0.00% | 4.6 |
| 215 | Qwen3.5-4B-heretic-imatrix-Q5_1.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.0237 | 3.0123 | 1.3938 | 2.9765 | 0.00% | 13.33% | 5.00% | 0.00% | 4.6 |
| 216 | Qwen_Qwen3.5-35B-A3B-IQ2_XS.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 3.7334 | 3.8853 | 0.9883 | 2.4180 | 0.00% | 13.33% | 5.00% | 0.00% | 4.6 |
| 217 | Qwen3.5-4B-IQ4_NL.gguf | unsloth/Qwen3.5-4B-GGUF | 5.5789 | 5.5816 | 0.4143 | 1.6132 | 0.00% | 8.33% | 10.00% | 0.00% | 4.6 |
| 218 | Qwen3.5-4B-heretic-imatrix-IQ4_NL.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 5.8234 | 5.8136 | 0.3625 | 1.5455 | 0.00% | 8.33% | 10.00% | 0.00% | 4.6 |
| 219 | Qwen_Qwen3.5-4B-Q4_K_L.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 4.3919 | 4.3886 | 0.6064 | 2.0492 | 0.00% | 8.33% | 10.00% | 0.00% | 4.6 |
| 220 | Qwen_Qwen3.5-4B-Q4_K_S.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 4.5126 | 4.5223 | 0.6083 | 1.9944 | 0.00% | 8.33% | 10.00% | 0.00% | 4.6 |
| 221 | Qwen_Qwen3.5-4B-Q5_K_L.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.7853 | 3.7723 | 0.9450 | 2.3776 | 0.00% | 8.33% | 10.00% | 0.00% | 4.6 |
| 222 | Qwen3.5-4B-UD-Q3_K_XL.gguf | unsloth/Qwen3.5-4B-GGUF | 3.6559 | 3.6483 | 0.9932 | 2.4618 | 0.00% | 6.67% | 10.00% | 0.00% | 4.2 |
| 223 | Qwen3.5-4B-heretic-imatrix-IQ4_XS.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.4717 | 4.4586 | 0.8055 | 2.0127 | 0.00% | 6.67% | 10.00% | 0.00% | 4.2 |
| 224 | Qwen3.5-4B-heretic-imatrix-Q5_K_S.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.6441 | 3.6489 | 1.0836 | 2.4697 | 0.00% | 11.67% | 5.00% | 0.00% | 4.2 |
| 225 | Qwen_Qwen3.5-4B-Q4_0.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 5.9310 | 5.9361 | 0.3174 | 1.5174 | 0.00% | 11.67% | 5.00% | 0.00% | 4.2 |
| 226 | Qwen3.5-0.8B-Q5_K_S.gguf | unsloth/Qwen3.5-0.8B-GGUF | 15.6990 | 15.7152 | 0.1841 | 0.5733 | 0.00% | 15.00% | 0.00% | 0.00% | 3.8 |
| 227 | Qwen3.5-0.8B-Q6_K.gguf | lmstudio-community/Qwen3.5-0.8B-GGUF | 15.4420 | 15.5455 | 0.1593 | 0.5829 | 0.00% | 15.00% | 0.00% | 0.00% | 3.8 |
| 228 | Qwen_Qwen3.5-0.8B-Q4_K_L.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 17.5616 | 17.5294 | 0.1240 | 0.5125 | 0.00% | 5.00% | 10.00% | 0.00% | 3.8 |
| 229 | \*Qwen_Qwen3.5-27B-IQ3_M.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.3339 | 0.3238 | 15.4135 | 26.9755 | 0.00% | 0.00% | 15.00% | 0.00% | 3.8 |
| 230 | \*Qwen_Qwen3.5-27B-Q3_K_L.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.0340 | 0.0349 | 19.6039 | 264.8257 | 0.00% | 0.00% | 15.00% | 0.00% | 3.8 |
| 231 | Qwen_Qwen3.5-35B-A3B-IQ3_M.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | NA | NA | NA | NA | 0.00% | 0.00% | 15.00% | 0.00% | 3.8 |
| 232 | Qwen_Qwen3.5-35B-A3B-Q3_K_XL.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | 2.6015 | 2.6154 | 3.5114 | 5.5182 | 0.00% | 0.00% | 15.00% | 0.00% | 3.8 |
| 233 | glm-4.7-flash-claude-4.5-opus.iq2_m.gguf | TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill-GGUF | 12.1268 | 10.9808 | 0.7110 | 1.7417 | 5.00% | 10.00% | 0.00% | 0.00% | 3.8 |
| 234 | Qwen3.5-4B.Q8_0.gguf | avalon2244/Qwen3.5-4B-Claude-Opus-4.6-Distilled-GGUF | 7.0059 | 6.9907 | 0.4629 | 1.9983 | 0.00% | 5.00% | 5.00% | 3.33% | 3.3 |
| 235 | Qwen_Qwen3.5-0.8B-Q3_K_XL.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 15.4153 | 15.4081 | 0.1983 | 0.5838 | 5.00% | 1.67% | 0.00% | 6.67% | 3.3 |
| 236 | Qwen_Qwen3.5-2B-Q4_1.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 9.6752 | 9.7521 | 0.3531 | 0.9302 | 5.00% | 6.67% | 0.00% | 1.67% | 3.3 |
| 237 | Qwen_Qwen3.5-4B-Q3_K_L.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.4114 | 3.4051 | 1.1814 | 2.6383 | 0.00% | 13.33% | 0.00% | 0.00% | 3.3 |
| 238 | \*Qwen3.5-4B-BF16.gguf | unsloth/Qwen3.5-4B-GGUF | 0.3419 | 0.3422 | 15.0289 | 26.3250 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 239 | Qwen3.5-4B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-4B-GGUF | 4.4576 | 4.4362 | 0.6157 | 2.0190 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 240 | Qwen3.5-4B-Q5_K_M.gguf | unsloth/Qwen3.5-4B-GGUF | 3.5641 | 3.5499 | 1.0352 | 2.5252 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 241 | Qwen3.5-4B-Q5_K_S.gguf | unsloth/Qwen3.5-4B-GGUF | 3.6190 | 3.6249 | 1.0752 | 2.4869 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 242 | Qwen3.5-4B-Q6_K.gguf | lmstudio-community/Qwen3.5-4B-GGUF | 3.5575 | 3.5141 | 0.8826 | 2.5299 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 243 | Qwen3.5-4B-Q8_0.gguf | lmstudio-community/Qwen3.5-4B-GGUF | 4.5077 | 4.4993 | 0.4610 | 1.9966 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 244 | Qwen3.5-4B-UD-Q8_K_XL.gguf | unsloth/Qwen3.5-4B-GGUF | 3.3198 | 3.3119 | 0.8685 | 2.7110 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 245 | Qwen3.5-4B-heretic-imatrix-Q4_K.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.4425 | 4.4389 | 0.6178 | 2.0259 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 246 | Qwen3.5-4B-heretic-imatrix-Q4_K_M.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.4324 | 4.4405 | 0.6199 | 2.0305 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 247 | Qwen3.5-4B-heretic-imatrix-Q5_0.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.3356 | 3.3357 | 1.2117 | 2.6982 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 248 | Qwen3.5-4B-heretic-imatrix-Q8_0.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 4.4663 | 4.4610 | 0.4811 | 2.0151 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 249 | Qwen_Qwen3.5-4B-Q4_K_M.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 4.3378 | 4.3114 | 0.6072 | 2.0748 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 250 | Qwen_Qwen3.5-4B-Q8_0.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 4.4341 | 4.4544 | 0.4919 | 2.0297 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 251 | \*Qwen_Qwen3.5-4B-bf16.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 0.3419 | 0.3417 | 15.0294 | 26.3224 | 0.00% | 8.33% | 5.00% | 0.00% | 3.3 |
| 252 | GLM-4.7-Flash-Uncen-Hrt-NEO-CODE-MAX-imat-D_AU-IQ2_M.gguf | DavidAU/GLM-4.7-Flash-Uncensored-Heretic-NEO-CODE-Imatrix-MAX-GGUF | 2.7161 | 3.0433 | 1.1977 | 3.0390 | 0.00% | 1.67% | 5.00% | 5.00% | 2.9 |
| 253 | Qwen3.5-4B-UD-Q4_K_XL.gguf | unsloth/Qwen3.5-4B-GGUF | 4.1500 | 4.1426 | 0.7496 | 2.1687 | 0.00% | 6.67% | 5.00% | 0.00% | 2.9 |
| 254 | Qwen3.5-4B.Q4_K_M.gguf | avalon2244/Qwen3.5-4B-Claude-Opus-4.6-Distilled-GGUF | 6.8811 | 6.8423 | 0.6215 | 2.0346 | 0.00% | 6.67% | 5.00% | 0.00% | 2.9 |
| 255 | Qwen_Qwen3.5-2B-Q4_K_M.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 10.1123 | 10.0916 | 0.2477 | 0.8900 | 0.00% | 6.67% | 0.00% | 5.00% | 2.9 |
| 256 | Qwen3.5-0.8B-Q4_K_S.gguf | unsloth/Qwen3.5-0.8B-GGUF | 18.1686 | 18.1507 | 0.1122 | 0.4954 | 0.00% | 11.67% | 0.00% | 0.00% | 2.9 |
| 257 | \*Qwen3.5-9B-abliterated-F16.gguf | lukey03/Qwen3.5-9B-abliterated-GGUF | 0.0273 | 0.0268 | 19.6192 | 329.9721 | 0.00% | 0.00% | 10.00% | 1.67% | 2.9 |
| 258 | Qwen_Qwen3.5-0.8B-Q5_K_S.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 15.6883 | 15.6632 | 0.1769 | 0.5737 | 0.00% | 5.00% | 5.00% | 1.67% | 2.9 |
| 259 | Qwen_Qwen3.5-35B-A3B-Q3_K_L.gguf | bartowski/Qwen_Qwen3.5-35B-A3B-GGUF | NA | 2.5063 | NA | NA | 0.00% | 11.67% | 0.00% | 0.00% | 2.9 |
| 260 | \*qwen2.5-coder-7b-instruct-fp16.gguf | Qwen/Qwen2.5-Coder-7B-Instruct-GGUF | 0.0854 | 0.0531 | 8.0098 | 185.6705 | 0.00% | 0.00% | 0.00% | 11.67% | 2.9 |
| 261 | \*Devstral-Small-2-24B-Instruct-2512-Q4_K_M.gguf | lmstudio-community/Devstral-Small-2-24B-Instruct-2512-GGUF | 0.0198 | 1.5483 | 65.3245 | 854.5761 | 0.00% | 5.00% | 0.00% | 5.00% | 2.5 |
| 262 | Qwen3.5-0.8B-Q4_1.gguf | unsloth/Qwen3.5-0.8B-GGUF | 16.6768 | 16.6628 | 0.1650 | 0.5397 | 0.00% | 10.00% | 0.00% | 0.00% | 2.5 |
| 263 | Qwen3.5-27B-Q2_K.gguf | unsloth/Qwen3.5-27B-GGUF | 0.7855 | 0.7921 | 5.8233 | 11.4612 | 0.00% | 0.00% | 10.00% | 0.00% | 2.5 |
| 264 | Qwen3.5-2B.i1-Q5_K_M.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 8.9769 | 8.9517 | 0.3851 | 1.0026 | 0.00% | 5.00% | 0.00% | 3.33% | 2.1 |
| 265 | Qwen_Qwen3.5-0.8B-Q4_K_S.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 17.4643 | 17.4729 | 0.1289 | 0.5153 | 0.00% | 3.33% | 5.00% | 0.00% | 2.1 |
| 266 | Qwen_Qwen3.5-2B-Q4_K_L.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 10.3829 | 10.3675 | 0.2463 | 0.8668 | 0.00% | 3.33% | 0.00% | 5.00% | 2.1 |
| 267 | Qwen3.5-4B-UD-Q5_K_XL.gguf | unsloth/Qwen3.5-4B-GGUF | 3.6768 | 3.6575 | 0.9316 | 2.4478 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 268 | Qwen3.5-4B-UD-Q6_K_XL.gguf | unsloth/Qwen3.5-4B-GGUF | 3.5266 | 3.5059 | 0.8845 | 2.5521 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 269 | \*Qwen3.5-4B-heretic-bf16.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 0.3418 | 0.3418 | 15.0280 | 26.3280 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 270 | Qwen3.5-4B-heretic-imatrix-Q5_K.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.5903 | 3.5836 | 1.0469 | 2.5068 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 271 | Qwen3.5-4B-heretic-imatrix-Q5_K_M.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.5894 | 3.5899 | 1.0448 | 2.5074 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 272 | Qwen3.5-4B-heretic-imatrix-Q6_K.gguf | ZuzeTt/Qwen3.5-4B-heretic-GGUF | 3.4670 | 3.4525 | 0.9112 | 2.5959 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 273 | Qwen_Qwen3.5-2B-Q5_K_L.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 9.2243 | 9.2927 | 0.3645 | 0.9757 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 274 | Qwen_Qwen3.5-4B-Q5_K_S.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.6179 | 3.6061 | 1.0409 | 2.4877 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 275 | Qwen_Qwen3.5-4B-Q6_K.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.6213 | 3.6015 | 0.8244 | 2.4853 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 276 | Qwen_Qwen3.5-4B-Q6_K_L.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.6661 | 3.6467 | 0.8205 | 2.4549 | 0.00% | 8.33% | 0.00% | 0.00% | 2.1 |
| 277 | Qwen3.5-0.8B-BF16.gguf | unsloth/Qwen3.5-0.8B-GGUF | 2.1332 | 2.1333 | 2.1142 | 4.2190 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 278 | Qwen3.5-2B-UD-Q6_K_XL.gguf | unsloth/Qwen3.5-2B-GGUF | 8.7132 | 8.7025 | 0.3367 | 1.0329 | 0.00% | 6.67% | 0.00% | 0.00% | 1.7 |
| 279 | Qwen3.5-2B.i1-Q4_1.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 9.9112 | 9.6941 | 0.3488 | 0.9081 | 0.00% | 3.33% | 0.00% | 3.33% | 1.7 |
| 280 | Qwen3.5-2B.i1-Q5_K_S.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 9.0481 | 9.0049 | 0.4040 | 0.9947 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 281 | Qwen3.5-4B.Q5_K_M.gguf | avalon2244/Qwen3.5-4B-Claude-Opus-4.6-Distilled-GGUF | 5.7110 | 5.6356 | 1.0125 | 2.4514 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 282 | Qwen_Qwen3.5-0.8B-Q4_0.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 20.5413 | 20.5115 | 0.0795 | 0.4381 | 0.00% | 3.33% | 0.00% | 3.33% | 1.7 |
| 283 | Qwen_Qwen3.5-0.8B-Q4_K_M.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 17.1638 | 17.1134 | 0.1257 | 0.5244 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 284 | Qwen_Qwen3.5-0.8B-Q8_0.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 17.9945 | 17.9921 | 0.1017 | 0.5002 | 0.00% | 3.33% | 0.00% | 3.33% | 1.7 |
| 285 | Qwen_Qwen3.5-0.8B-bf16.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 2.1328 | 2.1341 | 2.1136 | 4.2197 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 286 | Qwen_Qwen3.5-2B-Q5_K_S.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 8.6716 | 8.6549 | 0.4068 | 1.0379 | 0.00% | 5.00% | 0.00% | 1.67% | 1.7 |
| 287 | Qwen_Qwen3.5-2B-Q8_0.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 10.7016 | 10.7144 | 0.1920 | 0.8410 | 0.00% | 3.33% | 0.00% | 3.33% | 1.7 |
| 288 | Qwen_Qwen3.5-2B-bf16.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 0.8367 | 0.8365 | 5.7626 | 10.7570 | 0.00% | 3.33% | 0.00% | 3.33% | 1.7 |
| 289 | Qwen_Qwen3.5-4B-Q5_K_M.gguf | bartowski/Qwen_Qwen3.5-4B-GGUF | 3.7343 | 3.7208 | 0.9521 | 2.4101 | 0.00% | 6.67% | 0.00% | 0.00% | 1.7 |
| 290 | Qwen3.5-0.8B-UD-Q4_K_XL.gguf | unsloth/Qwen3.5-0.8B-GGUF | 16.7705 | 16.7623 | 0.1486 | 0.5367 | 0.00% | 1.67% | 0.00% | 3.33% | 1.2 |
| 291 | Qwen3.5-0.8B-UD-Q5_K_XL.gguf | unsloth/Qwen3.5-0.8B-GGUF | 15.9044 | 15.8697 | 0.1612 | 0.5659 | 0.00% | 0.00% | 5.00% | 0.00% | 1.2 |
| 292 | Qwen3.5-0.8B-UD-Q6_K_XL.gguf | unsloth/Qwen3.5-0.8B-GGUF | 15.7332 | 15.7297 | 0.1558 | 0.5720 | 0.00% | 3.33% | 0.00% | 1.67% | 1.2 |
| 293 | \*Qwen3.5-27B-UD-Q3_K_XL.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0457 | 0.0457 | 16.1796 | 196.8730 | 0.00% | 0.00% | 5.00% | 0.00% | 1.2 |
| 294 | Qwen3.5-2B-BF16.gguf | unsloth/Qwen3.5-2B-GGUF | 0.8362 | 0.8363 | 5.7644 | 10.7630 | 0.00% | 3.33% | 0.00% | 1.67% | 1.2 |
| 295 | Qwen3.5-2B-UD-Q8_K_XL.gguf | unsloth/Qwen3.5-2B-GGUF | 7.7325 | 7.7412 | 0.3437 | 1.1639 | 0.00% | 5.00% | 0.00% | 0.00% | 1.2 |
| 296 | Qwen3.5-2B.i1-IQ4_XS.gguf | mradermacher/Qwen3.5-2B-i1-GGUF | 10.6407 | 10.5942 | 0.3125 | 0.8458 | 0.00% | 3.33% | 0.00% | 1.67% | 1.2 |
| 297 | \*Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf | HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive | 0.0256 | 0.0247 | 42.6549 | 351.0031 | 0.00% | 0.00% | 5.00% | 0.00% | 1.2 |
| 298 | Qwen_Qwen3.5-0.8B-Q5_K_L.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 16.5032 | 16.5014 | 0.1585 | 0.5453 | 0.00% | 3.33% | 0.00% | 1.67% | 1.2 |
| 299 | Qwen_Qwen3.5-2B-Q5_K_M.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 9.0021 | 8.9996 | 0.3676 | 0.9998 | 0.00% | 3.33% | 0.00% | 1.67% | 1.2 |
| 300 | Qwen_Qwen3.5-2B-Q6_K.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 8.7229 | 8.7075 | 0.3226 | 1.0318 | 0.00% | 1.67% | 0.00% | 3.33% | 1.2 |
| 301 | \*qwen3.5-9b-null-space-abliterated-f16.gguf | jwest33/qwen3.5-9b-null-space-abliterated-GGUF | 0.0279 | 0.0268 | 20.6352 | 323.0082 | 0.00% | 0.00% | 5.00% | 0.00% | 1.2 |
| 302 | Qwen3.5-0.8B-UD-Q8_K_XL.gguf | unsloth/Qwen3.5-0.8B-GGUF | 14.1041 | 14.1174 | 0.1612 | 0.6381 | 0.00% | 1.67% | 0.00% | 1.67% | 0.8 |
| 303 | Qwen_Qwen3.5-0.8B-Q5_K_M.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 16.2036 | 16.1616 | 0.1599 | 0.5554 | 0.00% | 1.67% | 0.00% | 1.67% | 0.8 |
| 304 | Qwen_Qwen3.5-0.8B-Q6_K.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 15.9927 | 15.9524 | 0.1441 | 0.5628 | 0.00% | 3.33% | 0.00% | 0.00% | 0.8 |
| 305 | Qwen_Qwen3.5-0.8B-Q6_K_L.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 16.2512 | 16.2197 | 0.1437 | 0.5538 | 0.00% | 1.67% | 0.00% | 1.67% | 0.8 |
| 306 | Qwen_Qwen3.5-2B-Q6_K_L.gguf | bartowski/Qwen_Qwen3.5-2B-GGUF | 8.9115 | 8.8927 | 0.3205 | 1.0099 | 0.00% | 1.67% | 0.00% | 1.67% | 0.8 |
| 307 | huihui-ai_QwQ-32B-abliterated-Q2_K_L.gguf | bartowski/huihui-ai_QwQ-32B-abliterated-GGUF | 2.4004 | 2.2837 | 3.8431 | 8.5263 | 0.00% | 0.00% | 0.00% | 3.33% | 0.8 |
| 308 | Qwen3.5-2B-UD-Q4_K_XL.gguf | unsloth/Qwen3.5-2B-GGUF | 9.6567 | 9.6455 | 0.3134 | 0.9320 | 0.00% | 0.00% | 0.00% | 1.67% | 0.4 |
| 309 | Qwen_Qwen3.5-0.8B-Q4_1.gguf | bartowski/Qwen_Qwen3.5-0.8B-GGUF | 16.7115 | 16.6159 | 0.1623 | 0.5386 | 0.00% | 0.00% | 0.00% | 1.67% | 0.4 |
| 310 | \*Qwen3.5-27B-IQ4_XS.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0349 | 0.0347 | 19.7612 | 257.6839 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 311 | \*Qwen3.5-27B-Q4_1.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0311 | NA | 22.4319 | 289.6863 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 312 | \*Qwen3.5-27B-Q4_K_M.gguf | lmstudio-community/Qwen3.5-27B-GGUF | 0.0077 | 0.0091 | 66.9331 | 1162.7569 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 313 | \*Qwen3.5-27B-Q4_K_S.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0156 | 0.0148 | 34.8317 | 578.1193 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 314 | Qwen3.5-27B-UD-Q4_K_XL.gguf | unsloth/Qwen3.5-27B-GGUF | NA | NA | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 315 | Qwen3.5-27B-heretic-v2-Q4_K_M.gguf | llmfan46/Qwen3.5-27B-heretic-v2-GGUF | NA | NA | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 316 | \*Qwen3.5-27B-heretic-v2-Q4_K_S.gguf | llmfan46/Qwen3.5-27B-heretic-v2-GGUF | 0.0069 | 0.0069 | 97.3021 | 1303.9902 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 317 | \*Qwen3.5-27B-heretic-v2-Q5_K_M.gguf | llmfan46/Qwen3.5-27B-heretic-v2-GGUF | 0.0317 | 0.0317 | 22.9471 | 283.6233 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 318 | \*Qwen3.5-27B-heretic-v2-Q5_K_S.gguf | llmfan46/Qwen3.5-27B-heretic-v2-GGUF | 0.0281 | 0.0281 | 25.2277 | 320.6120 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 319 | Qwen3.5-2B-UD-Q5_K_XL.gguf | unsloth/Qwen3.5-2B-GGUF | 8.8078 | 8.7851 | 0.3616 | 1.0218 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 320 | Qwen3.5-35B-A3B-UD-IQ4_NL.gguf | unsloth/Qwen3.5-35B-A3B-GGUF | NA | 2.3455 | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 321 | Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive-Q3_K_M.gguf | HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive | NA | NA | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 322 | \*Qwen_Qwen3.5-27B-IQ4_NL.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.0122 | 0.0088 | 46.0507 | 740.7518 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 323 | \*Qwen_Qwen3.5-27B-IQ4_XS.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | 0.4016 | 0.2956 | 14.2061 | 22.8412 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 324 | Qwen_Qwen3.5-27B-Q3_K_XL.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | NA | NA | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 325 | Qwen_Qwen3.5-27B-Q4_0.gguf | bartowski/Qwen_Qwen3.5-27B-GGUF | NA | 0.0114 | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 326 | \*huihui-ai_QwQ-32B-abliterated-IQ3_M.gguf | bartowski/huihui-ai_QwQ-32B-abliterated-GGUF | 0.0662 | 0.0656 | 22.1170 | 256.9527 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 327 | phi-4-reasoning-vision-q8_0.gguf | jamesburton/Phi-4-reasoning-vision-15B-GGUF | NA | NA | NA | NA | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 328 | \*qwen2.5-coder-14b-instruct-q8_0.gguf | Qwen/Qwen2.5-Coder-14B-Instruct-GGUF | 0.0157 | 0.0111 | 58.8528 | 1115.8602 | 0.00% | 0.00% | 0.00% | 0.00% | 0.0 |
| 329 | \*Qwen3.5-27B-IQ4_NL.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0090 | 0.0093 | 52.4092 | 996.7356 | NA% | NA% | NA% | NA% | NA |
| 330 | \*Qwen3.5-27B-Q4_0.gguf | unsloth/Qwen3.5-27B-GGUF | 0.0090 | 0.0091 | 60.8902 | 1005.2170 | NA% | NA% | NA% | NA% | NA |
| 331 | \*Qwen3.5-27B-UD-IQ2_M.gguf | unsloth/Qwen3.5-27B-GGUF | 0.4898 | 0.4692 | 10.2128 | 18.4228 | NA% | NA% | NA% | NA% | NA |


## What's next

- **Higher-context KV cache testing** (8k, 16k, 32k) on the top 3 models to find the actual memory cliff
- **More model families** — suggestions welcome

## MLX autopilot run notes

This benchmark set also includes a full MLX run, executed with `mlx_lm.server` automation:

1. Start command:
```bash
export HF_TOKEN=your_token
export HF_OWNER=your_hf_username
bash scripts/start_mlx_autopilot-mlx.sh
```
2. Pipeline behavior:
- Starts `mlx_lm.server` per model (HF auto-download on first load)
- Runs perf sweep at ctx 1024/4096 and concurrency 1/2
- Runs quality (reasoning and non-reasoning)
- Uploads benchmark artifacts
- Deletes HF cache for tested model (`--delete-after-benchmark`)
3. Monitor:
```bash
tail -f ~/.mlx-autopilot/autopilot.log
cat ~/.mlx-autopilot/state.json
```

Final MLX outcome from this run:
- `43` candidates total
- `37` successful (`done`)
- `6` permanent failures
  - 3 VL repos (`LFM2.5-VL-1.6B-*`) are `mlx-vlm` models and fail under `mlx_lm.server`
  - 3 lmstudio 24B repos failed server startup (repo-side config issue)

Published MLX result files in this repo:
- `all_models_benchmark-mlx.csv` (includes `Status`)
- `all_models_benchmark_compat-mlx.csv` (same 28-column schema as the GGUF benchmark CSV)



## Consolidated MLX benchmark table (all 43 models)

Sorted by: `Status` (`done` first) then composite quality score (avg of R-GSM8K, R-MMLU, NR-GSM8K, NR-MMLU).

| # | Model | Source | Status | TPS (1k c1) | TPS (4k c1) | TTFT (s) | E2E (s) | R-GSM8K | R-MMLU | NR-GSM8K | NR-MMLU | Composite |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | LFM2-8B-A1B-qx64-hi-mlx | nightmedia | done | 26.1992 | 39.3898 | 0.4806 | 0.6981 | 60.00 | 50.00 | 35.00 | 50.00 | 48.8 |
| 2 | LFM2-8B-A1B-8bit-MLX | mlx-community | done | 15.7690 | 22.6677 | 0.6294 | 0.9214 | 50.00 | 51.67 | 30.00 | 55.00 | 46.7 |
| 3 | LFM2-8B-A1B-6bit-MLX | mlx-community | done | 24.6676 | 27.0987 | 0.3453 | 0.5778 | 55.00 | 48.33 | 25.00 | 55.00 | 45.8 |
| 4 | LFM2-8B-A1B-qx86-hi-mlx | nightmedia | done | 20.0318 | 21.9166 | 0.5561 | 0.8329 | 50.00 | 45.00 | 20.00 | 53.33 | 42.1 |
| 5 | LFM2-8B-A1B-mlx-mxfp4 | RepublicOfKorokke | done | 47.4267 | 56.4024 | 0.1583 | 0.3076 | 40.00 | 33.33 | 35.00 | 48.33 | 39.2 |
| 6 | LFM2-8B-A1B-4bit | mlx-community | done | 38.9098 | 49.1530 | 0.2169 | 0.3898 | 40.00 | 41.67 | 15.00 | 56.67 | 38.3 |
| 7 | LFM2-8B-A1B-3bit-MLX | mlx-community | done | 42.2475 | 50.3641 | 0.1764 | 0.3212 | 55.00 | 31.67 | 15.00 | 36.67 | 34.6 |
| 8 | LFM2-Deep-Horror-4B-mxfp4-mlx | nightmedia | done | 23.1837 | 25.5780 | 0.2002 | 0.5569 | 10.00 | 48.33 | 20.00 | 60.00 | 34.6 |
| 9 | LFM2-2.6B-4bit | mlx-community | done | 42.0452 | 44.0336 | 0.0947 | 0.3340 | 0.00 | 50.00 | 15.00 | 53.33 | 29.6 |
| 10 | LFM2.5-1.2B-Instruct-6bit | mlx-community | done | 79.4029 | 84.3909 | 0.0543 | 0.2153 | 25.00 | 35.00 | 15.00 | 35.00 | 27.5 |
| 11 | LFM2.5-1.2B-Instruct-MLX-6bit | LiquidAI | done | 70.2326 | 76.7488 | 0.0548 | 0.2131 | 25.00 | 35.00 | 15.00 | 35.00 | 27.5 |
| 12 | LFM2-2.6B-8bit | mlx-community | done | 22.3507 | 26.0905 | 0.1200 | 0.5574 | 5.00 | 50.00 | 0.00 | 53.33 | 27.1 |
| 13 | LFM2.5-1.2B-Instruct-8bit | mlx-community | done | 60.7967 | 60.3798 | 0.0571 | 0.2632 | 25.00 | 35.00 | 15.00 | 33.33 | 27.1 |
| 14 | LFM2.5-1.2B-Instruct-MLX-8bit | LiquidAI | done | 57.1592 | 60.3468 | 0.0576 | 0.2636 | 25.00 | 35.00 | 15.00 | 33.33 | 27.1 |
| 15 | LFM2-1.2B-dwq5-mlx | nightmedia | done | 57.8920 | 69.1603 | 0.0804 | 0.2291 | 10.00 | 31.67 | 15.00 | 46.67 | 25.8 |
| 16 | LFM2-1.2B-6bit | mlx-community | done | 56.7841 | 59.8874 | 0.0528 | 0.2136 | 5.00 | 31.67 | 15.00 | 50.00 | 25.4 |
| 17 | LFM2-1.2B-8bit | mlx-community | done | 51.8338 | 48.3959 | 0.0568 | 0.2622 | 5.00 | 31.67 | 20.00 | 45.00 | 25.4 |
| 18 | LFM2.5-1.2B-Instruct-4bit | mlx-community | done | 105.3347 | 113.9656 | 0.0496 | 0.1626 | 10.00 | 30.00 | 20.00 | 38.33 | 24.6 |
| 19 | LFM2.5-1.2B-Instruct-MLX-4bit | LiquidAI | done | 105.0193 | 113.1793 | 0.0498 | 0.1631 | 10.00 | 30.00 | 20.00 | 38.33 | 24.6 |
| 20 | LFM2-1.2B-5bit | mlx-community | done | 74.6321 | 79.6333 | 0.0512 | 0.1886 | 5.00 | 30.00 | 10.00 | 48.33 | 23.3 |
| 21 | LFM2-1.2B-dwq6-mlx | nightmedia | done | 42.8275 | 51.6714 | 0.0829 | 0.2550 | 5.00 | 28.33 | 10.00 | 46.67 | 22.5 |
| 22 | LFM2.5-1.2B-Instruct-5bit | mlx-community | done | 79.5112 | 84.6645 | 0.0532 | 0.1649 | 15.00 | 31.67 | 10.00 | 33.33 | 22.5 |
| 23 | LFM2.5-1.2B-Instruct-MLX-5bit | LiquidAI | done | 79.6493 | 83.5220 | 0.0527 | 0.1647 | 15.00 | 31.67 | 10.00 | 33.33 | 22.5 |
| 24 | LFM2-1.2B-4bit | mlx-community | done | 55.8437 | 60.1786 | 0.0494 | 0.1623 | 0.00 | 35.00 | 10.00 | 40.00 | 21.2 |
| 25 | LFM2.5-1.2B-Thinking-MLX-bf16 | LiquidAI | done | 24.6864 | 30.0492 | 0.1764 | 0.5441 | 5.00 | 8.33 | 10.00 | 20.00 | 10.8 |
| 26 | LFM2.5-1.2B-Thinking-8bit | mlx-community | done | 48.8757 | 53.9502 | 0.0607 | 0.2781 | 5.00 | 6.67 | 10.00 | 20.00 | 10.4 |
| 27 | LFM2.5-1.2B-Thinking-MLX-8bit | LiquidAI | done | 51.0287 | 54.5581 | 0.0879 | 0.2935 | 5.00 | 6.67 | 10.00 | 20.00 | 10.4 |
| 28 | LFM2.5-1.2B-Thinking-4bit | mlx-community | done | 92.2513 | 96.6658 | 0.0504 | 0.1639 | 0.00 | 6.67 | 15.00 | 16.67 | 9.6 |
| 29 | LFM2.5-1.2B-Thinking-MLX-4bit | LiquidAI | done | 82.8831 | 100.5385 | 0.0765 | 0.1900 | 0.00 | 6.67 | 15.00 | 16.67 | 9.6 |
| 30 | LFM2.5-1.2B-Thinking-5bit | mlx-community | done | 75.9109 | 79.2293 | 0.0542 | 0.1919 | 0.00 | 5.00 | 10.00 | 21.67 | 9.2 |
| 31 | LFM2.5-1.2B-Thinking-6bit | mlx-community | done | 63.2290 | 64.9148 | 0.0541 | 0.2153 | 5.00 | 6.67 | 5.00 | 20.00 | 9.2 |
| 32 | LFM2.5-1.2B-Thinking-MLX-5bit | LiquidAI | done | 67.0841 | 80.0430 | 0.0806 | 0.2178 | 0.00 | 5.00 | 10.00 | 21.67 | 9.2 |
| 33 | LFM2.5-1.2B-Thinking-MLX-6bit | LiquidAI | done | 55.6565 | 74.9011 | 0.0804 | 0.2411 | 5.00 | 6.67 | 5.00 | 20.00 | 9.2 |
| 34 | LFM2-24B-A2B-MLX-4bit | LiquidAI | done_no_perf* | NA | NA | NA | NA | 0.00 | 0.00 | 0.00 | 0.00 | 0.0 |
| 35 | LFM2-24B-A2B-MLX-5bit | LiquidAI | done_no_perf* | NA | NA | NA | NA | 0.00 | 0.00 | 0.00 | 0.00 | 0.0 |
| 36 | LFM2-24B-A2B-MLX-6bit | LiquidAI | done_no_perf* | NA | NA | NA | NA | 0.00 | 0.00 | 0.00 | 0.00 | 0.0 |
| 37 | LFM2-24B-A2B-mlx-mxfp4 | emberpadgett | done_no_perf* | NA | NA | NA | NA | 0.00 | 0.00 | 0.00 | 0.00 | 0.0 |
| 38 | LFM2-24B-A2B-MLX-4bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 39 | LFM2-24B-A2B-MLX-5bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 40 | LFM2-24B-A2B-MLX-6bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 41 | LFM2.5-VL-1.6B-MLX-4bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 42 | LFM2.5-VL-1.6B-MLX-6bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| 43 | LFM2.5-VL-1.6B-MLX-8bit | lmstudio-community | failed_permanent | NA | NA | NA | NA | NA | NA | NA | NA | NA |

\* `done_no_perf`: quality runs completed (`summary.tsv` present) but perf rows had no numeric metrics (`ok=0`), so latency/throughput are `NA`.



## The repo

Everything is open source — the autopilot script, benchmark harness, quality evaluation, data export, and analysis:

`https://huggingface.co/Manojb/macmini-16gb-bench-gguf-mlx`


To run your own benchmarks on your own hardware:
```bash
export HF_TOKEN=your_token
bash scripts/start_gguf_autopilot.sh
```

It handles downloading, benchmarking, quality evaluation, uploading, cleanup, crash recovery, and state resumption automatically. Add your own models by editing the wave definitions and restarting.

---

**Hardware**: Mac Mini M4, 16 GB unified memory (10 GPU cores, Metal 4), macOS 15.x, `llama-server` (llama.cpp)

**GPU offload note**: The autopilot now attempts Metal GPU acceleration (`-ngl auto`) first for every model. Models that fit in GPU memory (~12.4 GB working set) get full or partial GPU offload. Models that exceed GPU memory (27B+ dense with mmproj) automatically fall back to CPU-only (`-ngl 0`). All fallback events are logged. Models <= 14B and MoE models with small active params (e.g., 35B-A3B with 3B active) benefit from GPU acceleration. Earlier benchmark results (waves 1-33) were collected with `-ngl 0` (CPU-only); wave 34+ uses the GPU-first fallback.

**Methodology notes**: Quality eval uses compact subsets (20 GSM8K + 60 MMLU) — directionally useful for ranking but not publication-grade absolute numbers. Throughput numbers are p50 over multiple requests. All data is reproducible from the artifacts in the repo.