Instructions to use kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2")
model = AutoModelForMultimodalLM.from_pretrained("kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2

SGLang

How to use kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 with Docker Model Runner:
```
docker model run hf.co/kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2
```

Kataguru-Sceptic-35B-A3B-Basque-v3.2 — Full BF16 Master Weights (`254 + 2 = 256` MoE, `67.5 GB`)

Kataguru-Sceptic-35B-A3B-Basque-v3.2 is a sovereign Basque (Euskara) Mixture-of-Experts (254 + 2 = 256 experts per layer) language, reasoning, multimodal vision, and full-page translation model built on the Kataguru Sceptic 35B-A3B v3.1 foundation (15-Benchmark Mean: 70.26 %, GPQA Diamond: 70.20 %, LiveCodeBench v6: 39.43 %, BFCL V4 AST: 83.67 %).

Official Basque (`v3.2`) Release Suite

Full BF16 Master Weights (~67.5 GB): kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2
Blackwell Native NVFP4 (~24.3 GB): kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2-NVFP4
Universal vLLM W4A16-AWQ (~25.3 GB): kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2-W4A16-AWQ
GGUF APEX-Trio (APEX-Small, APEX-Medium, APEX-Good + mmproj): kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2-GGUF

1. What Kataguru Changed (`254 + 2` Sovereign Language MoE Architecture)

Field	Description / Value
Base model	`kataguru/Kataguru-Sceptic-35B-A3B-v3.1` (`254` frozen core experts `#0..#253`, `3B` active parameters per token)
Kataguru `v3.2` architecture	`254 + 2 = 256` Experts per MoE layer (`40` layers): • Experts `#0..#253` (`100 %` Frozen Core): Preserves the full scientific reasoning (`GPQA Diamond 70.20 %`), competitive coding (`LiveCodeBench v6 39.43 %`), agentic tool calling (`BFCL V4 83.67 %`), and industrial PLC/CNC capabilities with zero catastrophic forgetting. • Expert `#254` (Basque Grammar, Morphology & Technical Specialist): Ergative-absolutive case alignment, polypersonal auxiliary verb agreement (Nor-Nori-Nork), and agglutinative postpositional suffixes, plus industrial engineering, PLC/CNC, clinical medicine, and statutory/patent terminology. • Expert `#255` (Basque Literature, Nonfiction & Bidirectional Translator): Full-page literary prose, modern nonfiction, natural dialogue, and faithful `EN <-> EU` translation (`/translate`).
Language Bank Training Set	`4,844` human-verified golden samples (`100 %` Pass on Mechanical & Quality Gate) + contrastive router calibration (`--moe-mode new_experts_only --recalibrate-router`, `r=32, alpha=64`)
Dedicated Chat Template & Calibration	Native `Basque` `chat_template.jinja` (`/eu`, `/translate`, localized reasoning headers) + dedicated `Basque` quantization calibration (`calib_nvfp4_awq_256_eu.jsonl` & `imatrix_calibration_eu.txt`)
Format & Footprint	Full `BF16` Master Weights (`14` safetensors shards, `~67.5 GB`) — published openly for research, fine-tuning, and custom quantization
What was NOT changed	All `254` core reasoning/STEM/coding experts (`#0..#253`), `27`-layer multimodal vision encoder (`333` tensors), Layer 40 MTP speculative head (`785` tensors), `262,144` context window

2. Verified Benchmark & Live Audit Summary

Benchmark / Capability Pillar	Score	Notes
15-Benchmark Core Foundation Mean	70.26 %	Inherited 100 % intact via frozen Experts `#0..#253`
GPQA Diamond (PhD-level STEM)	70.20 % (`139 / 198`)	Graduate physics, chemistry & molecular biology
LiveCodeBench v6 (`Pass@1`)	39.43 % (`Easy: 90.70 %`)	Competitive algorithmic programming
BFCL V4 Single-Turn AST (Tool Calling)	83.67 % (`Non-Live: 88.44 %`)	Native `<tool_call>` AST & irrelevance detection (`80.26 %`)
Live Basque 10/10 Audit (`v3.2`)	10 / 10 PASS (`100 %`)	HiTZ EvalEU Suite (EusProficiency, EusExams, EusReading, EusTrivia) and Full-Page Literary & Technical Translation (EN/ES <-> EU)

3. Language-Specific Chat Template & Control Tags (`chat_template.jinja`)

Native Basque Mode: Prefix prompt with /eu (or pass chat_template_kwargs={"response_language": "eu"}).
Full-Page Translation Mode: Prefix prompt with /translate (sceptic_mode="translate") for zero-chatter, paragraph-preserving literary and technical translation between English, Finnish, and Basque.
Reasoning Depth: {REASON:spoon}, {REASON:einstein}, {REASON:007}, {REASON:deep}, {REASON:low}, or {REASON:off}.

4. Honest Limitations

High-Temperature Creative Fiction: For strict morphological and proofreading accuracy in Basque, use temperature=0.2–0.3 with presence_penalty=0.3.
VRAM Footprint: Full BF16 requires ~72 GB VRAM (2x 48–80 GB); NVFP4 / W4A16-AWQ runs on 2x RTX 5090 32GB (TP=2) or 1x 32 GB GPU; GGUF APEX-Medium (~13.6 GiB) runs on a single 16 GB GPU.

Laburpena (Euskaraz)

Kataguru-Sceptic-35B-A3B-Basque-v3.2 euskararako bereziki trebatutako 256 adituko MoE hizkuntza-eredua da (254 Sceptic oinarrizko aditu izoztu + 2 euskal aditu espezializatu #254 eta #255).

Suomenkielinen Yhteenveto / Finnish Summary

Kataguru-Sceptic-35B-A3B-Basque-v3.2 on Kataguru Sceptic 35B-A3B (v3.1, 15-benchmark: 70,26 %) -pohjalle rakennettu suvereeni 254 + 2 = 256 ekspertin kielimalli (Basque / Euskara):

Jäädytetty Sceptic-ydin (254 eksperttiä, #0..#253): Säilyttää 100 % monitoimimallin tieteellisestä päättelystä (GPQA Diamond 70,20 %), kilpailuohjelmoinnista (LiveCodeBench v6 39,43 %), työkalukutsuista (BFCL V4 83,67 %), teollisuusautomaatiosta (PLC/CNC) ja 27-kerroksisesta visiosta ilman unohtamista.
Kaksi dedikoitua kielieksperttiä (#254 ja #255, 4,844 näytettä):
- Ekspertti #254: Kohdekielen (Euskara) taivutusoppi, morfologia sekä tekniikan, automaation, lääketieteen ja lakikielen erikoissanasto.
- Ekspertti #255: Kohdekielen kaunokirjallisuus, tietokirjallisuus ja koko sivun kaksisuuntaiset käännökset (/translate).
Kielikohtainen kalibrointi ja täysi painojulkaisu: Mukana täydet BF16-painot, NVFP4, W4A16-AWQ sekä kielikohtaisella imatrix_calibration_eu.txt-korpuksella kalibroidut GGUF APEX-kvantit (APEX-Small, APEX-Medium, APEX-Good + mmproj-f16).

Downloads last month: -

Safetensors

Model size

36B params

Tensor type

BF16

Model tree for kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2

Base model

Qwen/Qwen3.5-35B-A3B-Base

Finetuned

Qwen/Qwen3.5-35B-A3B

Finetuned

kataguru/Kataguru-Sceptic-35B-A3B-v3.1

Finetuned

(10)

this model

Space using kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 1

Collection including kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2

Sovereign & Morphologically Complex Languages (254+2 MoE)

Collection

Sovereign LLMs for Finnish & morphologically complex languages (FI, ET, HU, LT, LV, IS, EU, KA, CY, TR) on Sceptic 35B-A3B 254+2 MoE. • 37 items • Updated about 3 hours ago

Kataguru-Sceptic-35B-A3B-Basque-v3.2 — Full BF16 Master Weights (254 + 2 = 256 MoE, 67.5 GB)

Official Basque (v3.2) Release Suite

1. What Kataguru Changed (254 + 2 Sovereign Language MoE Architecture)

2. Verified Benchmark & Live Audit Summary

3. Language-Specific Chat Template & Control Tags (chat_template.jinja)

4. Honest Limitations

Laburpena (Euskaraz)

Suomenkielinen Yhteenveto / Finnish Summary

Model tree for kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2

Space using kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2 1

Collection including kataguru/Kataguru-Sceptic-35B-A3B-Basque-v3.2

Kataguru-Sceptic-35B-A3B-Basque-v3.2 — Full BF16 Master Weights (`254 + 2 = 256` MoE, `67.5 GB`)

Official Basque (`v3.2`) Release Suite

1. What Kataguru Changed (`254 + 2` Sovereign Language MoE Architecture)

3. Language-Specific Chat Template & Control Tags (`chat_template.jinja`)