Gemini 3.1 Pro Preview | hokai.io
Gemini 3.1 Pro Preview: 1M context, 94.3% GPQA, reasoning model. $2/$12 per million tokens. Google DeepMind early access.
Google Gemini 3.1 Pro Preview: 1M context, 94.3% GPQA, advanced reasoning. Feb 2026 preview at $2/$12 per 1M tokens.
Gemini 3.1 Pro Preview is Google's preview reasoning model with 1M context and 94.3% GPQA Diamond. Released Feb 2026 at $2/$12 per 1M tokens.
Provider: Google · Family: Gemini 3.1
Context window: 1,000,000 tokens · Max output: 65,536
Input modalities: text, image, audio, video, pdf, tool-calls · Output: text, tool-calls
About Gemini 3.1 Pro Preview
Gemini 3.1 Pro Preview is Google DeepMind's early-access reasoning model released Feb 19, 2026. Identical to GA Gemini 3.1 Pro: sparse MoE Transformer 50-80B active params, 1M context, 65K output tokens. GPQA 94.3%, AIME 91.2%, MMLU-Pro 90.99%, SWE-bench 79.6%. Pricing: $2/$12 per 1M tokens. Multimodal: 900 images, 8.4 hours audio, 1 hour video input. Function calling, structured output, Search grounding, code execution, computer use all GA. Training cutoff Jan 2025. Safety: balanced, Constitutional AI + RLHF. System card at deepmind.google. Preview gets features 1-3 weeks before GA, 30-day deprecation vs 6-month for GA. Ideal for: research teams, early feature adopters. Not for: production needing long-term stability.
Benchmarks
- mmlu pro: 90.99
- aime 2025: 91.2
- gpqa diamond: 94.3
- swe bench verified: 79.6
Frequently Asked Questions
What is Gemini 3.1 Pro Preview?
Preview-tier Gemini 3.1 Pro variant released Feb 2026. Identical to GA: 1M context, 94.3% GPQA, sparse MoE architecture. $2/$12 pricing. Gets features 1-3 weeks earlier than GA. 30-day vs 6-month deprecation policy.