Qwen3.8 light tune
According to the model card, this is a light tune on Qwen3.8 using cited datasets plus a private Fable 5 corpus, with hub tags recording Qwen/Qwen3.8-27B as the base.
Open Source Model Profile · TeichAI
Qwen3.8-27B-Fable-Distill
Qwen3.8-27B-Fable-Distill is a 27.78B-parameter TeichAI tune of Qwen3.8-27B for image-text-to-text work. Its model card documents thinking controls and publisher-reported benchmark figures.
- Publisher
- TeichAI
- Task
- image-text-to-text
- Model type
- qwen3_5
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Qwen3.8-27B-Fable-Distill is published by TeichAI as an image-text-to-text model. The captured configuration identifies Qwen3_5ForConditionalGeneration with model type qwen3_5, and Safetensors metadata reports 27,781,427,952 parameters. According to the model card, it is a light tune on Qwen3.8, and hub tags record Qwen/Qwen3.8-27B as the base.
Recorded capabilities
Thinking controls
The model card documents enable_thinking with low, medium, or xhigh reasoning effort, plus base sampling recommendations of temperature 1.0, top_p 0.95, and top_k 20.
Publisher-reported benchmark table
The model card publishes third-party-credited ARC Challenge, ARC Easy, and BoolQ figures showing higher values for the distill than the base in that table.
Unsloth and TRL training stack
According to the model card, data work used Teich tooling and the qwen3_5 model trained with Unsloth and Hugging Face TRL, with BF16 weights and a chat template.
Use cases in the source record
- Conversational image-text-to-text experiments using the documented enable_thinking and reasoning-effort controls.
- Sampling-sensitive generation runs following the publisher's temperature 1.0, top_p 0.95, and top_k 20 recommendations.
Limitations and unknowns
- No context-window value was extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- Benchmark figures are publisher-reported third-party-credited values and were not independently verified by Ethen.
Source and provenance
Source: TeichAI/Qwen3.8-27B-Fable-Distill
Captured: Unknown. Processed: 2026-09-07T19:34:38.012731+00:00.
Qwen 3.8 Fable 5 Distill Light tune on Qwen3.8 on some the datasets cited here, as well as a large corpus of personal (private) Fable 5 data Model ARC Challenge ARC Challenge (Easy) BoolQ Qwen3.8-27B 0.591 0.782 0.896 Qwen3.8-27B-Fable-Distill 0.637 0.832 0.911 As always, big thank you to @nightmedia for the benchmarks Notes The model accepts enable_thinking and a reasoning_effort of low, medium or xhigh (the template's own default is xhigh, which thinks at length every turn). Base model sampling recommendations: temperature 1.0, top_p 0.95, top_k 20. The data for this model was easily formatted, validated, and masked using Teich Th…
F001F002F003F004F005F006F007F009F010F011F012F013F014F015