Gemini 3.1 Pro Preview
Google’s current Gemini flagship preview, tuned for better thinking, stronger software engineering behavior, and more grounded multimodal agent workflows.
Capability profile
- Input
- $2.00 / 1M tokens (200k)
- Output
- $12.00 / 1M tokens (200k)
- Cached input
- $0.20 / 1M tokens (200k)
- Context
- 1M tokens
- Free tier
- No / unknown
- API
- Available
- Subscription
- Gemini API, Google AI Studio, Vertex AI
- Open-source status
- Closed
- Large-context reasoning
- serious coding help
- multimodal analysis
- and agentic developer workflows.
- Repo-aware assistants
- multimodal agents
- long-context synthesis
- and coding-heavy product features.
- Ultra-cheap high-volume serving where Flash-class Gemini models make more sense.
- Products that only need low-cost lightweight inference or stable long-lived non-preview SKUs.
This looks like Google’s real flagship lane now: less about novelty, more about making Gemini feel dependable enough for serious agent and coding work.
Google’s current Gemini flagship preview, tuned for better thinking, stronger software engineering behavior, and more grounded multimodal agent workflows.
Post Reboot take
This looks like Google’s real flagship lane now: less about novelty, more about making Gemini feel dependable enough for serious agent and coding work.
Strengths
Best for: Large-context reasoning, serious coding help, multimodal analysis, and agentic developer workflows.
Weakest at: Ultra-cheap high-volume serving where Flash-class Gemini models make more sense.
Best workflows: Repo-aware assistants, multimodal agents, long-context synthesis, and coding-heavy product features.
Example use cases: Code review copilots, document-plus-image analysis, agent backends, and large-context technical workflows.