Google: Gemma 4 31B
◆ Strong7.8
/ 10
Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...
Specifications
| Attribute | Value |
|---|---|
| Lab | |
| Tags | Intelligent Open Weight Multimodal |
| Overall Score | 7.8/10 |
| Release Date | 2026-04 |
| Context Window | 262,144 tokens |
| Input Price / 1M | $0.12 |
| Output Price / 1M | $0.36 |
| Input Modalities | Image, Text, Video |
| Output Modalities | Text |
Strengths
- Strong 31B dense model with multimodal text+image input
- Configurable thinking/reasoning mode
- 256K context window for long-document work
- Open-weight with permissive license
Weaknesses
- Smaller than frontier models in total capability
- Limited agentic tool-use support
- Not as battle-tested as Gemini or other Google models
Best For
- Cost-effective multimodal applications
- Long-context document processing
- On-device and edge deployment
- Fine-tuning for specialized domains
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →