Gemma 4 31B
Language modelOpen weightsCard summarizedOur most intelligent open models to date, purpose-built for advanced reasoning and agentic workflows
Overview
Google · 02Gemma, front
Gemma 4 31B
Our most intelligent open models to date, purpose-built for advanced reasoning and agentic workflows
- Cost
- $0.20 host
- Context
- 256K
- Cheapest host
- $0.20 / $0.40
- Max output
- —
- Knows up to
- January 2025
- Released
- 2 Apr 2026
Gemma 4 is an Apache 2.0 open-weights family in five sizes (E2B, E4B, 12B Unified, 26B A4B MoE, 31B dense), released as pre-trained and instruction-tuned checkpoints. 31B is the largest: a 30.7B-parameter dense model with a 256K context that takes text and images and outputs text. Model information
gemma-4-31b-it- Input / output per 1M
- $0.20 / $0.40 via DeepInfra
- Context window
- 256K
- Max output
- —
- Knows up to
- January 2025
- Reasoning
- Configurable (<|think|> token in system prompt)
- Weights
- Open · Apache 2.0
About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Google DeepMind's Gemma 4 model card. Every claim carries its source page.