Skip to content
← Models
Open weightsGoogle/10 trustVerified
Google Gemma 4 31B logo

Google Gemma 4 31B

Gemma 4 flagship: 31B, thinking mode, native audio+vision, Apache 2.0.

Visit official page →

About

Released April 2, 2026, Gemma 4 31B is the flagship of Google's fourth open generation, built on Gemini 3 research. It adds a thinking mode for complex reasoning, native audio and vision multimodality (no separate encoder), a 256K context window, and reports 85.2% on MMMLU and 80% on LiveCodeBench. Notably, the whole Gemma 4 line moved to Apache 2.0, dropping the Gemma Terms of Use. The family spans E2B (2B) to 31B plus a 26B MoE.

Limitations: 31B dense is hardware-intensive for self-hosting; the line is young with limited production track record versus Gemma 3. Audio support is new and less battle-tested than the text/vision paths.

Specs

Context window

256K tokens

Pricing

Free weights (Apache 2.0); API via Google AI Studio

Free quota

Free via AI Studio, Kaggle, Hugging Face

API access

Hugging Face weights, Google AI Studio, Kaggle, Vertex AI

Weights

Open weights

Released

April 2, 2026

Benchmarks

Team-reported: 85.2% MMMLU, 80% LiveCodeBench

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

reasoningmultimodalaudioagents

Trust score — 9/10

/10 trustVerified

Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 44 (models); Trending score 86/100 from researcher assessment.

More in the gemma family

Sources