AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF

Parable-Qwen3-4B-Claude-Fable-5-GGUF

[image]

A 4B local coding model with agent instincts. Planning, tool habits and terminal reasoning distilled from real Claude Fable 5 agent sessions, not synthetic Q&A. Runs on ~2.5 GB of RAM.

ollama run hf.co/AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF:Q4_K_M

v3.1 (2026-08-11)

Retrained on corpus v3.1: the v2 agent traces plus 1,807 execution-verified solutions generated by the previous build and kept only where the code actually ran against its tests. Two seeds souped, merged at the v2.1 scale.

The result matches or beats base Qwen3-4B on all four execution benchmarks, where the previous build trailed it on three. If you pulled this model before 11 August 2026, re-pull.

Files

What it is good at

• —It answers. Base Qwen3-4B spends its whole budget inside <think> on 34% of ordinary prompts and returns nothing. This model answers 34/34 on the same suite, with 140x less reasoning text and no thinking-mode flag to manage.
• —Agent-shaped reasoning. Trained on genuine multi-step agent sessions, so plans, tool selection and terminal workflows come out structured instead of improvised.
• —Small enough to keep open. Q4KM is 2.5 GB. Laptop, old GPU, modest desktop — it runs, offline, with your code staying on your machine.

Evaluation

Measured on identical harnesses, greedy decoding, Q4KM builds, thinking disabled on every row. Base and this model run through the same instrument in the same session.

Measured on the v2.1 build and carried forward (the training objective and chat behaviour are unchanged):

Choosing between this and the base

Take this model for local agent and coding work where you want structured, reliable answers every time: it fits the agent-session distribution far better and never silently returns empty.

Take the base model if your workload is maximum-accuracy function calling in a tool-calling harness, where its few extra points matter more than reasoning style.

Model details

• —Base: Qwen/Qwen3-4B (4B, Apache-2.0)
• —Method: QLoRA (nf4, r16, alpha 32) on all-linear targets, completion-only loss masking, 30% general-instruction replay mix, seed-averaged weights, merged at scale 0.6 (v2.1 recalibration)
• —Data: genuine Claude Fable 5 agent sessions + gpt5.5-terminal transcripts, deduplicated and decontaminated against the reported benchmarks
• —Method report: doi:10.5281/zenodo.21676407

Provenance & licensing

Fine-tuned from Qwen/Qwen3-4B (Apache-2.0). Training data: Glint-Research/Fable-5-traces (AGPL-3.0) and Roman1111111/gpt5.5-terminal (MIT). Because those traces originate from third-party assistants, the providers' terms may apply to downstream training and distillation. If you plan to build on this model commercially, confirm your use aligns with those terms.

Citation

@misc{aglawe2026agenttrace,
  author    = {Aglawe, Ankit},
  title     = {Agent-Trace Fine-Tuning of Small Language Models under Constrained Compute},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.21676407},
  url       = {https://doi.org/10.5281/zenodo.21676407}
}

Acknowledgements

The Qwen team for the base model; Glint-Research and Roman1111111 for the trace datasets; empero-ai for the recipe this series iterates on.