Gemini 全部可用模型（实时）

Gemini 3 Pro in preview. Frontier reasoning plus full multimodal input — the most capable Gemini available right now, with the usual preview caveats.

Preview of Gemini 3's fast tier. Sub-second responses with frontier-grade knowledge — the drop-in successor to 2.5-flash once GA ships.

Next-generation Gemini Pro in preview. Track frontier capability gains before GA — expect occasional behavior changes between snapshots.

Preview of Gemini 3.1 at the cheapest tier. Early bird for high-volume routing before GA lands.

Cheapest Gemini 3.1 variant. For trivial classification, lightweight rewrites, and embeddings-adjacent batch work.

Forward-looking flash preview. Earliest peek at the Gemini 3.5 lineage — useful for benchmarking what's coming, not for production.

Google's production flagship. 1M-token context, deep reasoning, native vision + audio + video — the default pick for long-document analysis and multimodal workloads.

Google's workhorse. 1M context at the industry's lowest per-token rate — ideal for retrieval-heavy apps and high-volume routing.

Cheapest current Gemini. For trivial classification, lightweight rewrites, and large-batch summarization where quality isn't paramount.

Stable 2.0 baseline. Pinned for legacy code paths — migrate to 2.5-flash for any new work.

Lite version of 2.0 Flash. Pinned for legacy budget workloads.