skip to content
gigarouter gigarouter
models / text generation · coming soon

Gemma 4 12B Coder

yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF

published Jun 2026 · updated Jun 2026

Gemma 4 12B Coder is a text-generation model that generates Python code solutions from chain-of-thought reasoning, fine-tuned from Google's Gemma 4 12B IT.

status
coming soon
API providers
0
downloads / mo
628.2K
license
apache-2.0

specs

TaskText Generation (Code)
ArchitectureGemma 4 (Transformer)
Parameters12B
LicenseApache 2.0

about this model

gemma-4-12B-coder-fable5-composer2.5-v1-GGUF is a text-generation model fine-tuned from Google's Gemma 4 12B for verifiable Python coding tasks. It uses chain-of-thought reasoning distilled from two complementary sources: genuine reasoning traces from Composer 2.5 that were verified by running code against test suites, and synthetic "second-attempt" traces from Fable 5 for problems the first source solved incorrectly. All training examples were gated on passing the task's tests, ensuring the emitted code is runnable.

The model supports a native context window of 256,000 tokens (262,144) after correcting an upstream metadata bug. It retains Gemma 4's thinking mode and is optimized for Python algorithmic problems. Key characteristics include reduced refusals compared to the base model, as the training data is task-focused without safety hedging. Developers should add their own guardrails for production use. The model is English-centric and general-knowledge accuracy should be verified independently.

The model is released under the Apache 2.0 license, inheriting from Gemma 4's licensing, and is free to use, modify, and redistribute.

best for

FAQ

What is the maximum context length?

256K tokens (262,144).

What license does this model use?

Apache 2.0, free to use, modify, and redistribute.

How can I call this model via the gigarouter API?

Use the gigarouter OpenAI-compatible endpoint with your API key.

What GGUF quantization sizes are available?

Q2_K (4.5 GB), Q3_K_M (5.7 GB), Q4_K_M (6.87 GB), Q6_K (9.11 GB), and Q8_0 (11.8 GB).

Is this model safety-aligned for production?

No, it has reduced refusals and is not safety-aligned; add your own guardrails.

not yet live

We're benchmarking and onboarding Gemma 4 12B Coder as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.

related text generation models

compare all →