group string | family string | author string | quant string | candidate string | mean_kld float64 | expert_cell_kld float64 | routing_term float64 | weakest_domain string | weakest_domain_kld float64 | median_kld float64 | p99_kld float64 | max_kld float64 | top1_agreement float64 | positions int64 | law_compliant bool |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ceccb250f2fb | Qwen3.6-27B | Qwen | FP8 | Qwen3.6-27B/Qwen3.6-27B-FP8 | 0.01601 | null | null | dialogue_instruction | 0.084519 | 0.001775 | 0.092164 | 31.782639 | 0.968365 | 1,572,096 | true |
25575e44a52a | Qwen3.6-35B-A3B | Qwen | FP8 | Qwen3.6-35B-A3B/Qwen3.6-35B-A3B-FP8 | 0.018381 | 0.009267 | 0.012317 | dialogue_instruction | 0.045155 | 0.003926 | 0.204983 | 24.641069 | 0.953085 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | Qwen | FP8 | Qwen3.8-27B/Qwen3.8-27B-FP8 | 0.011047 | null | null | dialogue_instruction | 0.053887 | 0.0019 | 0.086461 | 24.250298 | 0.967078 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | cyankiwi | AWQ-INT4 | Qwen3.8-27B/Qwen3.8-27B-AWQ-INT4 | 0.031102 | null | null | dialogue_instruction | 0.096677 | 0.006866 | 0.375823 | 28.042185 | 0.938757 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | unsloth | NVFP4 | Qwen3.8-27B/Qwen3.8-27B-NVFP4 | 0.040969 | null | null | dialogue_instruction | 0.124578 | 0.009442 | 0.505674 | 36.545666 | 0.92606 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | dbirks | W4A16-AutoRound | Qwen3.8-27B/Qwen3.8-27B-W4A16-AutoRound | 0.042758 | null | null | dialogue_instruction | 0.128057 | 0.008979 | 0.557369 | 28.063667 | 0.927893 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | RadixArk | NVFP4-BF16-LMHead | Qwen3.8-27B/Qwen3.8-27B-NVFP4-BF16-LMHead | 0.05213 | null | null | dialogue_instruction | 0.145061 | 0.01303 | 0.671704 | 27.482958 | 0.915187 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | philbert440 | W4A16-AWQ | Qwen3.8-27B/Qwen3.8-27B-W4A16-AWQ | 0.05214 | null | null | dialogue_instruction | 0.159012 | 0.01289 | 0.651413 | 30.399612 | 0.918826 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | RedHatAI | INT4 | Qwen3.8-27B/Qwen3.8-27B-INT4 | 0.062979 | null | null | dialogue_instruction | 0.144605 | 0.013248 | 0.928208 | 31.918503 | 0.911585 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | RadixArk | NVFP4 | Qwen3.8-27B/RadixArk-Qwen3.8-27B-NVFP4 | 0.064629 | null | null | dialogue_instruction | 0.153104 | 0.023806 | 0.696257 | 27.040554 | 0.892982 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | gittensor-model-hub | NVFP4-RTX5090 | Qwen3.8-27B/Qwen3.8-27B-NVFP4-RTX5090 | 0.088849 | null | null | dialogue_instruction | 0.212686 | 0.031539 | 1.040555 | 29.621611 | 0.879409 | 1,572,096 | true |
93196c409645 | Qwen3.8-27B | Inferact | NVFP4 | Qwen3.8-27B/Inferact-Qwen3.8-27B-NVFP4 | 0.138753 | null | null | dialogue_instruction | 0.322449 | 0.044459 | 1.767216 | 29.817425 | 0.859725 | 1,572,096 | true |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Local Inference Lab: distribution-fidelity index
Every published one-pager, with a pointer to the full artifact that backs it. Each artifact repo carries the token suite and the reusable reference tensors, so a third party can score a new candidate without loading the reference checkpoint.
Read LAWS.md before comparing anything here. Numbers are comparable only within a single artifact's suite, geometry, and runtime identity.
Qwen3.6-27B
Artifact: phaedawg/qwen3.6-27b-distribution-fidelity-768x2048-v1
- Qwen FP8 β mean KLD 0.01601047
Qwen3.6-35B-A3B
Artifact: phaedawg/qwen3.6-35b-a3b-distribution-fidelity-768x2048-v1
- Qwen FP8 β mean KLD 0.01838144
Qwen3.8-27B
Artifact: phaedawg/qwen3.8-27b-distribution-fidelity-768x2048-v1
- Qwen FP8 β mean KLD 0.01104734
- cyankiwi AWQ-INT4 β mean KLD 0.03110211
- unsloth NVFP4 β mean KLD 0.04096872
- dbirks W4A16-AutoRound β mean KLD 0.04275781
- RadixArk NVFP4-BF16-LMHead β mean KLD 0.05212951
- philbert440 W4A16-AWQ β mean KLD 0.05213981
- RedHatAI INT4 β mean KLD 0.06297880
- RadixArk NVFP4 β mean KLD 0.06462909
- gittensor-model-hub NVFP4-RTX5090 β mean KLD 0.08884902
- Inferact NVFP4 β mean KLD 0.13875295
- Downloads last month
- 46