Google releases Gemma 4 31B with 256K context and configurable reasoning mode
Google DeepMind has released Gemma 4 31B, a 30.7-billion-parameter multimodal model supporting text and image input. The model features a 262,144-token context window, configurable thinking/reasoning mode, native function calling, and multilingual support across 140+ languages under Apache 2.0 license.
Gemma 4 31B at a glance
How Gemma 4 31B ranks now
| Benchmark | Score | Rank | Source |
|---|---|---|---|
| Arena Elo | 1452 Elo | 53 of 141 | LMArena, Nov 2026 ↗ |
| GPQA | 85.7% | 78 of 163 | Artificial Analysis, Jul 2026 ↗ |
| Hallucination (Summarization) (lower is better) | 7.4% | 14 of 49 | Vectara, Nov 2026 ↗ |
| MMLU-Pro | 85.2% | 25 of 74 | llm-stats, Oct 2026 ↗ |
| LiveCodeBench | 80.0% | 20 of 48 | Hugging Face model card ↗ |
Rank is against every model in our database with a sourced score on that benchmark, each counted once with its best score.
Google Releases Gemma 4 31B Multimodal Model
Google DeepMind has released Gemma 4 31B, a 30.7-billion-parameter dense multimodal model designed for both text and image input processing. The model launched on April 2, 2026, and is available through OpenRouter at $0.14 per million input tokens and $0.40 per million output tokens.
Key Specifications
The Gemma 4 31B Instruct variant includes a 256,144-token context window—among the largest for models in its parameter class. The model supports configurable thinking/reasoning mode, enabling step-by-step reasoning for complex tasks. It includes native function calling capabilities and multilingual support across 140+ languages.
Google is releasing the model under the Apache 2.0 open license, allowing commercial and research use with minimal restrictions.
Capabilities and Performance
According to Google DeepMind, Gemma 4 31B demonstrates particular strength in three areas: coding tasks, reasoning-heavy problems, and document understanding. The configurable reasoning mode allows developers to trade latency for reasoning depth—enabling the model to show its internal thought process before producing final answers.
The multimodal architecture supports both text and image input, though the model outputs text only. This positions it as a document analysis and visual question-answering tool.
What This Means
Gemma 4 31B enters a crowded market of open 30B-class models from Meta (Llama), Mistral, and others, but differentiates on three fronts: the massive 256K context window (useful for long document processing), the explicit reasoning mode (reflecting broader industry trend toward chain-of-thought capabilities), and the Apache 2.0 license (lowest legal friction for commercial deployment).
The pricing—$0.14/$0.40 input/output—is competitive with similar-scale open models. The 256K context window is particularly notable; it enables processing of entire codebases or lengthy documents in a single request, reducing the need for context management and retrieval systems.
For organizations deploying locally or on proprietary infrastructure, the open-source weights and permissive license remove API dependency concerns. The 30.7B parameter count positions it as deployable on consumer-grade hardware (though requiring 60GB+ VRAM for full precision).
Related Articles
Qwen4 Max launches with 1M-token context and $1.50/$4.50 per 1M token pricing
Qwen4 Max, the flagship model in Alibaba's Qwen4 series, is listed on OpenRouter with a 1M-token context window. It accepts text, image and video input, returns text, and is priced at $1.50 per 1M input tokens and $4.50 per 1M output tokens.
Runway releases Praxis-1 robot model with open weights under Apache 2.0, tops World Action Models on RoboLab-120
Runway has released Praxis-1, a robot control model, with open weights and source code under the Apache 2.0 license. It scores 46.2 percent on Nvidia's RoboLab-120 leaderboard, ahead of FLUX 3 Action at 42.9 percent, in tests Runway ran itself.
Meituan releases LongCat-2.5: 1.6T-parameter multimodal MoE with 48B active per token
Meituan's LongCat team has published LongCat-2.5, a native multimodal mixture-of-experts model with 1.6 trillion total parameters and approximately 48 billion activated per token. It targets long-horizon agentic tasks such as codebase development and full-project builds, and ships with W4A4 expert quantization for domestic AI accelerators.
inclusionAI releases Duo 1.0 causal-base, a full-duplex multimodal world model initialized from MiniMax H3
inclusionAI has published the Duo 1.0 causal-base weights on Hugging Face. Duo is described as a full-duplex multimodal world model for real-time audio-visual interaction, initialized from MiniMax H3 and trained further by the Duo team.
Comments
Loading...