模型与价格
15 个模型，一个 Key。
官方价，全部划掉。
按请求计费，按下方价格扣费：无订阅、永不过期。点开任意一行查看模型 ID、支持接口和每一项价格。
模型与价格
claude-haiku-5-5提示超过 100,000 tokens 按更高档位计费。
官方价来源: basellm.github.io
打开 Playground →Nano Banana Pro: high-quality image generation and editing for complex layouts, precise typography and professional creative work. Inputs: text and images. Outputs: images and text. Input limit: 65,536 tokens; output limit: 32,768. Supports images up to 4K. Use native Gemini generateContent with IMAGE in responseModalities. Text output and image output are priced separately.
gemini-3-pro-imageclaude-opus-5-5Nano Banana 2.1: the recommended image generation and editing model for design, text rendering and consistent multi-turn edits. Inputs: text, images, video and PDF. Outputs: images and text. Input limit: 131,072 tokens; output limit: 32,768. Supports 1K, 2K and 4K images. Use the native Gemini generateContent endpoint with IMAGE in responseModalities; do not use OpenAI /v1/images/generations.
gemini-nano-banana-2.1Gemini 3.8 Flash TTS: high-fidelity text-to-speech for narration, expressive voices and longer audio. Input: text. Output: audio. Input limit: 8,192 tokens; output limit: 16,384. Follow the response mimeType for the audio format. Use native Gemini generateContent with responseModalities=["AUDIO"] and speechConfig. This gateway does not support this model through OpenAI /v1/audio/speech. Promotional prices end on December 31, 2026 and double on January 1, 2027.
gemini-3.8-flash-ttsclaude-fable-5-1claude-sonnet-5-5Gemini 3.8 Flash: the recommended general-purpose model for multimodal understanding, coding, reasoning and responsive applications. Inputs: text, images, video, audio and PDF. Output: text. Input limit: 1,048,576 tokens; output limit: 65,536. Supports Chat Completions, Responses and Gemini generateContent/streamGenerateContent. Current promotional prices end on December 31, 2026; base input/output prices double on January 1, 2027. The minimal thinking level is not supported.
gemini-3.8-flashGemini 3.5 Flash-Lite: a low-cost, low-latency option for high-volume classification, translation, extraction and document processing. Inputs: text, images, video, audio and PDF. Output: text. Input limit: 1,048,576 tokens; output limit: 65,536. Supports Chat Completions, Responses and Gemini generateContent/streamGenerateContent.
gemini-3.5-flash-liteGemini 3.1 Pro Preview: the Pro preview option for complex coding, careful reasoning and tool-driven workflows. Inputs: text, images, video, audio and PDF. Output: text. Input limit: 1,048,576 tokens; output limit: 65,536. Supports Chat Completions, Responses and Gemini generateContent/streamGenerateContent. No shutdown date has been announced. Prompts over 200K tokens use the higher long-context pricing tier.
gemini-3.1-pro-preview提示超过 200,000 tokens 按更高档位计费。
官方价来源: aistudio.google.com
打开 Playground →Gemini 3.8 Flash-Lite TTS: low-cost, low-latency text-to-speech for interactive playback and high-throughput speech generation. Input: text. Output: audio. Input limit: 8,192 tokens; output limit: 16,384. Follow the response mimeType for the audio format. Use native Gemini generateContent with responseModalities=["AUDIO"] and speechConfig. OpenAI /v1/audio/speech is not supported by this gateway. Promotional prices end on December 31, 2026 and double on January 1, 2027.
gemini-3.8-flash-lite-ttsGemini Embedding 001: a mature text embedding model for existing RAG, semantic search and similarity applications. Input: plain text. Output: an embedding vector. Input limit: 2,048 tokens. Supports custom outputDimensionality through the dimensions parameter; recommended dimensions are 768, 1536 and 3072. Use the OpenAI-compatible /v1/embeddings endpoint.
gemini-embedding-001Gemini 4 Argon: flagship reasoning for demanding coding, mathematics, analysis and multi-step agents. Inputs: text, images, video, audio and PDF. Output: text. Unified input/output context: 1,048,576 tokens; standard synchronous output limit: 65,536. Available through Chat Completions, Responses and Gemini generateContent/streamGenerateContent. Current promotional prices end on November 30, 2026; base input/output prices double on December 1 and require a pricing update.
gemini-4-argonNano Banana 2 Lite: economical image generation and editing for interactive and high-volume visual workflows. Inputs: text, images, video and PDF. Outputs: images and text. Input limit: 65,536 tokens; output limit: 4,096. Only 1K images are supported, not 2K or 4K. Use native Gemini generateContent with IMAGE in responseModalities.
gemini-3.1-flash-lite-imageGemini Embedding 2: next-generation embeddings for semantic search, RAG, clustering and similarity matching. The gateway currently exposes plain-text input through /v1/embeddings. Input limit: 8,192 tokens. Output: an embedding vector, using the upstream default of 3,072 dimensions. Although the Google model is multimodal, this gateway does not expose multimodal embedding inputs or forward the dimensions parameter for this model.
gemini-embedding-2下表价格为每百万 tokens（USD）。控制台里能看到每次请求的实际扣费。
计算器