hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF · Hugging Face

Source: https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#%F0%9F%94%A5-qwen36-35b-a3b-claude-46-opus-reasoning-distilled-gguf🔥 Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF

GGUF quantizations ofhesamation/Qwen3\.6\-35B\-A3B\-Claude\-4\.6\-Opus\-Reasoning\-Distilled, a reasoning SFT fine-tune ofQwen/Qwen3\.6\-35B\-A3Bon Claude Opus 4.6-style chain-of-thought distillation data.

The source fine-tune is text-only. The Qwen3.6 base architecture includes a vision encoder, but this fine-tuning run did not train on image or video examples. Treat these GGUF files as text-generation/runtime quantizations of the merged fine-tuned checkpoint.

• Developed by:@hesamation
• Source model:hesamation/Qwen3\.6\-35B\-A3B\-Claude\-4\.6\-Opus\-Reasoning\-Distilled
• Base model:Qwen/Qwen3\.6\-35B\-A3B
• **License:**apache-2.0

This fine-tuning run is inspired byJackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled, including the notebook/training workflow style and Claude Opus reasoning-distillation direction.

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#available-gguf-quantizationsAvailable GGUF Quantizations

This repo is intended to host the following GGUF variants. Files are uploaded as each quantization finishes.

QuantTypical useQ4\_K\_MSmallest practical general-purpose quant for local inferenceQ5\_K\_MBetter quality/size balance than Q4Q6\_KHigher-quality quant when VRAM/RAM budget allowsQ8\_0Largest quant here; closest to source quality among these options

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#benchmark-resultsBenchmark Results

The benchmark below was run on the merged source model, not separately on each GGUF quant. Quantization can change scores, especially at lower bitrates, so treat this as source-checkpoint context.

The MMLU-Pro pass used 70 total questions per model:\-\-limit 5across 14 MMLU-Pro subjects. Treat this as a smoke/comparative check, not a release-quality full benchmark.

BenchmarkHarnessSamples per modelSettingMetricBase modelSource merged modelDeltaMMLU-Pro overalllm-evaluation-harness70\-\-limit 5across 14 subjectsexact_match, custom-extract42.86%75.71%+32.85 pp Base model:Qwen/Qwen3\.6\-35B\-A3B. Source merged model:hesamation/Qwen3\.6\-35B\-A3B\-Claude\-4\.6\-Opus\-Reasoning\-Distilled.

Community benchmarks welcome To better understand this fine-tuned model and its GGUF quantizations, I welcome independent benchmark results. If you run evaluations, please include the benchmark name, harness/script, sample count, decoding settings, quant file, and raw logs or result files when possible. Share results by opening a PR/discussion or DMing@hesamationon X.

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#training-summaryTraining Summary

Qwen/Qwen3.6-35B-A3B
  -> supervised fine-tuning with LoRA
  -> merged full model
  -> GGUF quantization with llama.cpp

SettingValueFine-tuning methodSupervised fine-tuning with LoRALoRA targetAttention-only modulesLoRA rank / alpha32 / 32Micro-batch size1Gradient accumulation32Epochs2Completed steps762 / 762Final reported training loss0.3362497625740494Dataset max tokens8192Max sequence length32768

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#training-dataTraining Data

The source model samples and normalizes reasoning conversations from three datasets, then renders them with theqwen3\-thinkingchat template and response-only SFT masking.

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#intended-useIntended Use

These GGUF files are intended for local or server-side text inference through runtimes that support GGUF and the Qwen3.6 architecture, such as recentllama\.cppbuilds. Choose the quantization based on your memory budget and quality target.

Because the fine-tune is text-only, image/video behavior should be treated as inherited from the base model rather than improved by this training run.

https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GGUF#acknowledgementsAcknowledgements

Thanks to the Qwen team for the base model,Unslothfor the training stack,llama.cppfor GGUF tooling, andJackrongfor the public reasoning-distillation workflow that inspired this fine-tune.