Model Detail

gemma4-31B-Fable-5-Distilled is a code generation model with 31B parameters released by autotrust. The model is registered under the image-text-to-text pipeline tag on Hugging Face, released under the gemma license.

gemma4-31B-Fable-5-Distilled ships with 31B parameters. Total weight footprint is approximately 31.3 GB, which is the relevant figure when planning local-inference VRAM. Distribution is governed by the gemma license — review the exact terms before commercial deployment.

gemma4-31B-Fable-5-Distilled is best fit for code completion, repository-scale Q&A, and pair-programming integrations. It is a less obvious choice for one-shot generation of security-critical code without review. Treat this as a starting matrix rather than a benchmark verdict — the right deployment usually depends on the specific evaluation suite that mirrors your workload.