GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
rico03
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
rico03
Model Tree
Qwen/Qwen3.8-27B
Input Modalities
Output Modalities
Supported Functionality
Qwen3.8-27B is natively vision-language with a hybrid attention stack (16/64 full-attention layers, 48/64 Gated DeltaNet/linear-attention layers). This fine-tune targets only the text reasoning path — both attention types via LoRA, MLP layers, vision tower left frozen and unused.
train_on_responses_only)| Steps trained | 150 (partial run — pipeline validation, not full convergence) |
| Epochs | 0.126 (~12.6% of one epoch, ~2,700 of 21,490 examples) |
| Effective batch size | 18 (per-device 6 × grad accum 3) |
| Final train loss (avg) | 0.728 |
| Final eval loss | not computed this run (eval_strategy="no" to save time) |
| Wall-clock time | 53 min on 1× H100 NVL |
| Sequence length | 8192 (99.7% of dataset uncut; 0.3% longest examples excluded rather than truncated) |
| Trainable params | 233,455,616 (0.85% of 27.6B) |
| Adapter size on disk | 954 MB |
Status: pipeline-validation run, not a fully converged model. 150 steps at batch 18 covers ~12.6% of one epoch — enough to confirm the training pipeline (LoRA on hybrid attention, masking, chat template, checkpointing) works end-to-end, not enough for the model to have generalized across the full dataset. A full run (1+ epoch, ~1,194+ steps) is needed before this should be treated as a finished distillation. Loss trended downward over the run (high ~0.95-1.1 early, several sub-0.6 readings by step 70-90) but with a cosine schedule calibrated for exactly 150 steps — this checkpoint cannot simply be resumed for a longer run; a full run needs a fresh schedule sized to the full dataset.
| Source | Examples | What it is |
|---|---|---|
| lordx64/reasoning-distill-claude-opus-4-7-max | 8,124 | Genuine Claude Opus 4.7 extended-thinking traces (verified model/usage fields per row) |
| Jackrong/Claude-opus-4.7-TraceInversion-5000x | 4,800 | Real Opus 4.7 outputs (teacher_model field), reasoning trace reconstructed post-hoc by a smaller "trace inversion" model — plausible CoT matching a genuine answer, not captured original thinking |
| Jackrong/Claude-opus-4.6-TraceInversion-9000x |
Transparency note: ~62% of the dataset (the two TraceInversion sets) has a genuine final answer from Opus but a reconstructed reasoning trace, not Opus's actual internal thinking. This is disclosed so downstream users know the reasoning style is a plausible approximation for roughly two-thirds of the data, not a verbatim mirror throughout. The lordx64 subset (~38%) is the closest to authentic Opus thinking style.
We deliberately excluded several other "Claude-distilled" datasets found on Hugging Face after inspection — some contained self-referential synthetic text describing itself as "exemplifying" a model's style rather than real model output, and others were low-effort duplicate repos farming downloads. Always verify teacher_model/usage provenance fields before trusting a distillation dataset.
These are official Qwen3.8-27B numbers for the base model, included for reference — this fine-tune has not been independently re-benchmarked yet.
| Qwen3.8-27B | Qwen3.6-27B | Qwen3.7-Plus | |
|---|---|---|---|
| Terminal-Bench 2.1 (agentic coding) | 73.0 | 63.4 | 64.0 |
| SWE-bench Pro | 61.7 | 53.5 | 57.6 |
| QwenSWEBench | 79.0 | 49.3 | 59.2 |
| CoWorkBench (long-horizon office work) | 70.7 | 61.0 | 65.1 |
| Qwen3.8-27B | Qwen3.6-27B | Qwen3.7-Plus | |
|---|---|---|---|
| OSWorld-Verified (computer use) | 84.3 | 63.9 | 73.3 |
| AndroidWorld (mobile use) | 81.9 | 70.3 | 81.0 |
| MathVision (with CI) | 94.6 | 85.1 | 90.3 |
| OmniDocBench 1.5 | 91.1 | 89.4 |
Full base model benchmark tables: Qwen/Qwen3.8-27B model card.
python
from unsloth import FastLanguageModelmodel, tokenizer = FastLanguageModel.from_pretrained(model_name = "rico03/Qwen3.8-27B-Claude-Opus-Reasoning-Distilled",max_seq_length = 8192,dtype = None,load_in_4bit = True,)FastLanguageModel.for_inference(model)messages = [{"role": "user", "content": "Explain the difference between TCP and UDP."}]inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to("cuda")outputs = model.generate(inputs, max_new_tokens=1024, temperature=1.0, top_p=0.95, top_k=20)print(tokenizer.decode(outputs[0]))
Thinking is on by default (<think>...</think> block before the final answer), matching the base Qwen3.8 chat template. Use reasoning_content as a separate message field if constructing multi-turn history manually — do not pack the thinking block into content, the chat template expects it separate.
<think> format and some stylistic shift, but not robust generalization across task domains. A full training run is needed for production use — see status note above.Training methodology based on the Jackrong fine-tuning guide. Thanks to lordx64 and Jackrong for the source reasoning datasets.
Maintained by rico03
| 8,700 |
| Same trace-inversion method, Opus 4.6 |
| IFBench (instruction following) | 79.5 | 69.1 | 79.1 |
| GPQA Diamond (scientific reasoning) | 89.2 | 87.8 | 90.3 |
| HLE (multidisciplinary reasoning) | 30.8 | 24.0 | 34.7 |
| LiveCodeBench v6 | 90.3 | 83.9 | 89.6 |