--- license: apache-2.0 base_model: Qwen/Qwen3.8-27B library_name: transformers pipeline_tag: image-text-to-text tags: - qwen3.8 - reasoning - vision-language - personal-model - uncensored - abliterated - abliterix - transformers - bfloat16 --- # Qwen3.8-27B-Opus-Abliterix-Reasoning-BF16 This is the full-precision BF16 form of the selected Abliterix pass-1 practical winner. It preserves the checkpoint before MLX quantization and exists for Transformers workflows, archival fidelity, and downstream conversion. **Selection verdict:** practical personal-model winner with measured deviations — not universal dominance and not a claim that all preregistered gates passed. ## Why this release The target was a practical personal reasoning/VLM model with much lower reflexive refusal while retaining measured capability and an immutable fallback. The benchmark table is the decision record: it shows where the selected winner improved, where the control stayed stronger, and why both are published. The family therefore describes the winner as a **practical personal-model selection with measured deviations**, not universal dominance. ## What this variant is - **Role:** Full-precision practical winner - **Format:** Merged BF16 Transformers checkpoint with the native 15-tensor MTP sidecar preserved as `mtp.safetensors`. - **Base:** [`Qwen/Qwen3.8-27B`](https://huggingface.co/Qwen/Qwen3.8-27B) - **Pipeline:** image + text to text - **License:** Apache-2.0 ## Frozen local results > **These are self-run, frozen local project benchmarks, not official Qwen benchmarks.** The same local harness compared `control-bf16` with `abliterix-pass1-bf16`. | Frozen local metric | Control | Abliterix winner | |---|---:|---:| | Harmful hard refusal | 43.2% | **0.0%** | | Harmful soft deflection | 14.6% | **0.2%** | | Harmful substantive response | 47.0% | **99.4%** | | Capability macro | 17.6859% | **21.0086%** | | Full code | **16/421** | 10/421 | | HumanEval | **7.9268%** | 4.2683% | | Long-form pass | 54.1667% | **62.5000%** | | MMMU30 | 9/30 | **11/30** | | Held-out loss ratio | 1.000000 | 1.024478 | | Benign KL | 0.000000 | 0.093614 | The winner improved the local capability macro by 3.3227 points, long-form pass rate by 8.3333 points, and MMMU30 by 2 correct answers while reducing harmful hard refusal by 43.2 points. The control remained better on full code/HumanEval. Full machine-readable values are in [`benchmark-results.json`](./benchmark-results.json). ## Strict deviations and code pathology - Benign KL was `0.093614`, above the strict `0.05` limit. - Incoherence was `4.3077%`, above the `2.7692%` strict comparison point. - HumanEval was `4.2683%` versus control `7.9268%`; full code was `10/421` versus `16/421`. - Long-form maximum repeated-4gram fraction was `5.8632%`, above the `5%` limit. - Prompt leakage was detected: 3 exact winner prompt echoes versus 2 for control; both failed the leakage hard gate. - 376/421 winner code generations hit the 512-token cap. Among generations that reached execution, the winner passed 10/46 (21.74%) versus control 16/103 (15.53%), pointing to severe termination/extraction pathology rather than a clean latent-code estimate. ## Training data and run | Aggregate source label | Rows | |---|---:| | `opus-10000x` | 9,633 | | `opus-3000x` | 2,326 | | `reasoning-700x` | 633 | | `high-reasoning-250x` | 250 | | **Raw total** | **12,842** | After 208 deduplications and 20 invalid-row removals, 12,614 rows remained: 12,349 train, 127 validation, and 138 test. The processed-manifest SHA-256 is `6e0a36ad20732c5f98ff592c4565a4c86876fead4e9f94bc6ceedfad1339a94d`. Only aggregate counts and hashes are published; raw/private rows are not. Training ran for 1,544 optimizer steps with 108,789,760 trainable LoRA parameters. Final validation loss was `0.23739749`; token accuracy was `91.7594%`. ## Architecture and lineage The model uses `Qwen3_5ForConditionalGeneration` as recorded by the released config: a 64-layer, hidden-size-5120 text stack with a 3:1 linear/full-attention schedule and a configured 262,144-token maximum position range, paired with a 27-layer, hidden-size-1152 vision encoder. It is an image/text conditional-generation model, not a text-only checkpoint. ### Method 1. Started from [`Qwen/Qwen3.8-27B`](https://huggingface.co/Qwen/Qwen3.8-27B), Apache-2.0. 2. Prepared 12,842 raw reasoning rows: 12,614 accepted after removing 208 duplicates and 20 invalid rows, then split into 12,349 train / 127 validation / 138 test rows. 3. Trained a reasoning QLoRA on the 12,349-row train split for 1,544 optimizer steps with 108,789,760 trainable LoRA parameters. Final validation loss was 0.23739749 and token accuracy was 91.7594%. 4. Merged the adapter to BF16 to form immutable `control-bf16`. 5. Produced the selected winner with Abliterix 1.12.2 pass 1 (seed 42): orthogonal/projected, winsorized single-direction residual steering over the output/down-projection writer components. Q/K/V projections were excluded. 6. Preserved BF16 releases and converted both variants to affine MLX 8-bit, group size 64. Native one-layer MTP drafters were split and validated separately. The final selection label is `selected_practical_winner_with_measured_deviations`. Private training data, raw harmful/benign prompt sets, operational receipts, local paths, and Drive metadata are intentionally not published. ## Tensor and conversion integrity Across the BF16 comparison there were 1,199 tensor keys, including 15 native MTP tensors and 333 vision tensors. The winner contains 74 actual Abliterix residual-writer edits and zero unexpected changes. MLX native proof ran through `mlx-vlm` on arm64 macOS/Metal. Ordinary and MTP-assisted generation both produced the exact answer `323` for winner and control. The winner MLX build used `mlx` 0.32.0, `mlx-lm` 0.31.3, `mlx-vlm` 0.6.13, and affine 8-bit group-64 quantization. ## Hash and size summary - `BF16 winner delivery`: 14 files, 55,583,144,390 bytes, sealed aggregate SHA-256 `cb0e04180dc257f19604f9ae70b65191a9f01322e5ed7c36aecf22a00af173be` The deliberately published model payload for this repository is 10 files / 55,583,125,224 bytes. Per-file source SHA-256 values are in [`SHA256SUMS`](./SHA256SUMS) and [`manifests/artifact-manifest.json`](./manifests/artifact-manifest.json). Sealed aggregate hashes above cover the complete local source components, including private entries excluded from publication; use the per-file public manifest for the Hub payload. ## Use with Transformers A current Transformers version that supports Qwen3.8/Qwen3.5 VLM architecture is required (the sealed build used 5.15.0). BF16 needs substantial accelerator memory. ```python import torch from transformers import AutoModelForImageTextToText, AutoProcessor repo_id = "timteh673/Qwen3.8-27B-Opus-Abliterix-Reasoning-BF16" processor = AutoProcessor.from_pretrained(repo_id) model = AutoModelForImageTextToText.from_pretrained( repo_id, torch_dtype=torch.bfloat16, device_map="auto" ) messages = [{"role": "user", "content": [ {"type": "text", "text": "Calculate 17 multiplied by 19."} ]}] inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt" ).to(model.device) output = model.generate(**inputs, max_new_tokens=256) trimmed = output[:, inputs.input_ids.shape[1]:] print(processor.batch_decode(trimmed, skip_special_tokens=True)[0]) ``` For image input, add an image content item using the upstream Qwen multimodal message format. `mtp.safetensors` is the preserved native MTP sidecar; ordinary Transformers loading does not require it. ## Known limitations - The winner is deliberately less refusal-prone. The frozen harmful benchmark measured 0.0% hard refusal, not a guarantee of zero refusal on every prompt. - It failed strict KL, incoherence, HumanEval/full-code, repetition, and prompt-leakage gates listed above. - Code results are heavily confounded by 512-token termination; increase generation limits and validate executable outputs. - BF16 is large and requires substantial memory; MTP support depends on the runtime. - The local benchmark suite and private training/evaluation data are not included. Results may not transfer to other prompts, languages, runtimes, or sampling settings. - Generated content can be incorrect, insecure, or incomplete; review it before consequential use. ## Release family - [Recommended MLX 8-bit winner](https://huggingface.co/timteh673/Qwen3.8-27B-Opus-Abliterix-Reasoning-MLX-8bit) - [BF16 winner](https://huggingface.co/timteh673/Qwen3.8-27B-Opus-Abliterix-Reasoning-BF16) - [MLX 8-bit control](https://huggingface.co/timteh673/Qwen3.8-27B-Opus-Reasoning-Control-MLX-8bit) - [BF16 control](https://huggingface.co/timteh673/Qwen3.8-27B-Opus-Reasoning-Control-BF16) See [`RELEASE_FAMILY.md`](./RELEASE_FAMILY.md) for the role and integrity summary of every variant. ## Responsible use Evaluate this model for your own setting, isolate untrusted code/tool output, and comply with applicable law and the policies of systems you connect it to. The uncensoring/abliteration work changes refusal behavior; it does not make outputs accurate or safe by default. ## License and attribution Released under Apache License 2.0. This is a derivative of [`Qwen/Qwen3.8-27B`](https://huggingface.co/Qwen/Qwen3.8-27B), Copyright 2026 Alibaba Cloud. Qwen/Alibaba Cloud attribution and the full license text are retained in [`LICENSE`](./LICENSE). Modifications include the personal reasoning QLoRA merge, selected Abliterix residual edits for winner variants, and MLX quantization where applicable. ## Support If this release is useful, follow [`timteh673`](https://huggingface.co/timteh673), star the repository, and share reproducible benchmark or runtime findings in the Hub community tab.