{"v":1,"id":"model:hf:yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","slug":"model-yuxinlu1-gemma-4-12b-coder-fable5-composer2-5-v1-gguf","kind":"model","category":"llm","title":"gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","summary":"💻 Gemma4-12B-Coder (GGUF) — Composer 2.5 × Fable 5 ✨ ### 🐣 Tiny footprint, big brain — a local coding model for everyone","source":{"provider":"hf","ref":"yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","url":"https://huggingface.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","rev":"1380be1796e559fca96b4107599285cab3ddbb92","fetchedAt":"2026-10-02T20:53:38.403Z","etag":"W/\"5332-SKzuxnMK90svHEJjJAuLHqoCYrU\""},"author":{"name":"yuxinlu1","url":"https://huggingface.co/yuxinlu1"},"license":{"spdx":"apache-2.0","raw":"apache-2.0","open":true},"metrics":{"downloads":405736,"downloadsWeek":327838,"likes":2919,"takenAt":"2026-10-02T20:53:38.403Z"},"tags":["gguf","gemma4","coding","code","reasoning","thinking","llama.cpp","local-llm","text-generation","endpoints_compatible","conversational"],"pipeline":"text-generation","links":{"github":"google-deepmind/gemma","npm":"node-llama-cpp"},"updatedAt":"2026-06-19T03:52:11.000Z","collectedAt":"2026-10-02T20:53:38.403Z","review":{"numbers":["405,736 downloads on Hugging Face","2,919 likes","license apache-2.0","6.9 GB for gemma4-coding-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":"2026-10-02T20:53:38.403Z","http":200},"description":"# 💻 Gemma4-12B-Coder (GGUF) — Composer 2.5 × Fable 5 ✨\n### 🐣 Tiny footprint, big brain — a local **coding** model for *everyone*\n\n> **No matter your GPU. No matter your RAM.** If you've got **~4.5 GB** of VRAM *or* unified memory free,\n> you can run your own private, offline coding assistant right now. 🚀\n> This is the **v1 / code edition** — distilled from **real chain-of-thought** so it *thinks through* a problem\n> before writing the solution. 🧠💻 All local, all yours, no API, no cloud.\n\n### 🎯 What it is\nA focused fine-tune of Gemma 4 12B on **verifiable Python coding** data — every training example's reasoning leads to\ncode that **actually passed its tests**. The result reasons in the open (edge cases, complexity, approach) and then\nemits a clean, runnable solution. 💚\n\n---\n\n## 📌 Announcements\n\n**🚀🔥 IT'S HERE — v2 is OUT NOW!** v2 has shipped — the **GGUF quants are live and ready to run** →\n**[grab v2 here](https://huggingface.co/yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF)**. 🎉\nThe **full `safetensors` master** (build / fine-tune on top) goes up **tomorrow**. v2 is **agentic + coding** focused —\nthe piece v1 was missing.\n\n**Here's the result that got me most excited.** When I saw v2's **tau2-bench `telecom`** result — an agentic tool-use\nbenchmark where the model has to *diagnose → fix → verify*, exactly like real terminal/debugging work — I literally got\n**launched out of my chair** (…okay, *kidding* 😄). The jump in **actually solving the problem** is wild:\n\n| tau2-bench **telecom** · local, same harness, **Q8_0** | score |\n|---|---|\n| official `gemma-4-12B-it` (base) | **~15%** |\n| 🟢 **v2 (this release)** | **~55%** |\n\nThe base model tends to **give up early** (hands the problem off to a human); **v2 keeps going** and works it the way a\nmuch bigger model would. Full benchmark details are in the **[v2 card](https://huggingface.co/yuxinlu1/gemm…\n\nSource: https://huggingface.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","install":{"kind":"model","hfId":"yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF","gated":false,"format":"gguf","files":[{"name":"gemma4-coding-Q2_K.gguf","size":4830147104,"quant":"Q2_K","sha256":"adbd331a74121b78f223b40b3a51f4418aa6f55122c1fa3711b0615f56bf8729"},{"name":"gemma4-coding-Q3_K_M.gguf","size":6087086624,"quant":"Q3_K_M","sha256":"b2d2ca4e1fc592a7ba77225032224938b9f159ef6431f57e1db1b5e612f50b54"},{"name":"gemma4-coding-Q4_K_M.gguf","size":7381381664,"quant":"Q4_K_M","sha256":"1fe90b72e105d7bc71650aa59883edece3e84751af489075217a7ae717b1fe8d"},{"name":"gemma4-coding-Q6_K.gguf","size":9786020384,"quant":"Q6_K","sha256":"9a4aa42ef5c540afedaeec47194c39abc9268745662dc44b8351f658e979b4b1"},{"name":"gemma4-coding-Q8_0.gguf","size":12669645344,"quant":"Q8_0","sha256":"18629e26f7b800357fe95ae3804c9be49af58ebc73e80754c301ebe997e29fbb"}],"totalBytes":40754281120,"suggestedFile":"gemma4-coding-Q4_K_M.gguf","requirements":{"ramGb":9,"diskBytes":7381381664,"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:yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF"}}