{"v":1,"id":"model:hf:AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF","slug":"model-ankitai-parable-qwen3-4b-claude-fable-5-gguf","kind":"model","category":"llm","title":"Parable-Qwen3-4B-Claude-Fable-5-GGUF","summary":"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.","source":{"provider":"hf","ref":"AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF","url":"https://huggingface.co/AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF","rev":"abe4021465bae1921f1d172b49ad817702fb69a7","fetchedAt":"2026-10-02T21:00:58.556Z","etag":"W/\"1d0d-hVOIyBqJ6TqyqRGzRR77iLCeASA\""},"author":{"name":"AnkitAI","url":"https://huggingface.co/AnkitAI"},"license":{"spdx":"apache-2.0","raw":"apache-2.0","open":true},"metrics":{"downloads":348992,"downloadsWeek":327838,"likes":19,"takenAt":"2026-10-02T21:00:58.556Z"},"tags":["llama.cpp","gguf","qlora","agentic","coding","reasoning","qwen3","local-llm","ollama","lm-studio","text-generation","dataset:Glint-Research/Fable-5-traces","dataset:Roman1111111/gpt5.5-terminal","endpoints_compatible","conversational"],"pipeline":"text-generation","links":{"github":"QwenLM/Qwen3","npm":"node-llama-cpp"},"updatedAt":"2026-09-22T12:40:55.000Z","collectedAt":"2026-10-02T21:00:58.556Z","review":{"numbers":["348,992 downloads on Hugging Face","19 likes","license apache-2.0","2.3 GB for Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q4_K_M.gguf","327,838 npm downloads a week for node-llama-cpp","latest node-llama-cpp@3.22.1"],"log":null},"trust":"unlabeled","health":{"status":"alive","checkedAt":"2026-10-02T21:00:58.556Z","http":200},"description":"# Parable-Qwen3-4B-Claude-Fable-5-GGUF\n\n**A 4B local coding model with agent instincts.** Planning, tool habits and\nterminal reasoning distilled from real Claude Fable 5 agent sessions, not\nsynthetic Q&A. Runs on ~2.5 GB of RAM.\n\n```bash\nollama run hf.co/AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF:Q4_K_M\n```\n\n## v3.1 (2026-08-11)\n\nRetrained on **corpus v3.1**: the v2 agent traces plus 1,807 execution-verified\nsolutions generated by the previous build and kept only where the code actually\nran against its tests. Two seeds souped, merged at the v2.1 scale.\n\nThe result **matches or beats base Qwen3-4B on all four execution benchmarks**,\nwhere the previous build trailed it on three. If you pulled this model before\n11 August 2026, re-pull.\n\n## Files\n\n| File | Quant | Size | |\n|---|---|---|---|\n| Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q4_K_M.gguf | Q4_K_M | 2.5 GB | **recommended** |\n| Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q5_K_M.gguf | Q5_K_M | 2.9 GB | |\n| Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q6_K.gguf | Q6_K | 3.3 GB | |\n| Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q8_0.gguf | Q8_0 | 4.3 GB | |\n| Parable-Qwen3-4B-Claude-Fable-5-GGUF-F16.gguf | F16 | 8.1 GB | for re-quantizing |\n\n## What it is good at\n\n- **It answers.** Base Qwen3-4B spends its whole budget inside `` on\n  34% of ordinary prompts and returns nothing. This model answers 34/34 on\n  the same suite, with 140x less reasoning text and no thinking-mode flag to\n  manage.\n- **Agent-shaped reasoning.** Trained on genuine multi-step agent sessions,\n  so plans, tool selection and terminal workflows come out structured\n  instead of improvised.\n- **Small enough to keep open.** Q4_K_M is 2.5 GB. Laptop, old GPU, modest\n  desktop — it runs, offline, with your code staying on your machine.\n\n## Evaluation\n\nMeasured on identical harnesses, greedy decoding, Q4_K_M builds, thinking\ndisabled on every row. Base and this model run through the same instrument i…\n\nSource: https://huggingface.co/AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF","install":{"kind":"model","hfId":"AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF","gated":false,"format":"gguf","files":[{"name":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-F16.gguf","size":8051284800,"quant":"F16","sha256":"f6c2220ab470c6f012965888dde3f9e1a8bc3d829bcb64b872d8d81d8e185f7e"},{"name":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q4_K_M.gguf","size":2497280320,"quant":"Q4_K_M","sha256":"4b94ae1f56142eb04dc54f810fece7bc904a8ff855169b8ec4e57a366c9eabcc"},{"name":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q5_K_M.gguf","size":2889513280,"quant":"Q5_K_M","sha256":"76200de852de38fa9ddc11c1a13c95e567f1f8d889c41d22ee3d3eaf5333674d"},{"name":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q6_K.gguf","size":3306260800,"quant":"Q6_K","sha256":"936f508ac1dfc93adc53d396b42bbb73622ddae7f6df8194ee650236fa069cbe"},{"name":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q8_0.gguf","size":4280404800,"quant":"Q8_0","sha256":"7b9b99a25c3511bea7a16ebf2c5dc83188e885b5a7e75a914a90eacb684d75fd"}],"totalBytes":21024744000,"suggestedFile":"Parable-Qwen3-4B-Claude-Fable-5-GGUF-Q4_K_M.gguf","requirements":{"ramGb":4,"diskBytes":2497280320,"note":"estimate: suggested file size × 1.15 + 0.5 GB; a real measurement comes with lsh models"},"runWith":["llama.cpp","ollama"],"command":"lsh models install hf:AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF"}}