An Anthropic model for coding, long-context knowledge work, and visual understanding with adaptive thinking.
Use anthropic/claude-opus-5-5 through the sjolt LLM API.
anthropic/claude-opus-5-5Token rates
anthropic/claude-opus-5-5Similar Models
A fast DeepSeek model for text generation and reasoning tasks.
A DeepSeek model for text generation and reasoning tasks that need more capability.
Google’s fast reasoning model for agentic coding and complex knowledge workflows.
Model overview
Analyze code, reason over long documents, and interpret images with adaptive thinking through the native Anthropic Messages API.
Provide relevant files, the intended behavior, and clear acceptance criteria. Ask for a diagnosis, a focused implementation, or a review grounded in the supplied code.
Work with up to 1,000,000 context tokens. Label source sections and state the desired output so the model can connect evidence across documents and conversation turns.
Combine image inputs with specific questions about diagrams, screenshots, or visual details. Outputs are text; the illustrations on this page are conceptual artwork.
Model characteristics
Up to 1,000,000 context tokens and 128,000 output tokens. Set max_tokens to leave room for both thinking and the final response.
Thinking stays enabled. Select low, medium, high, xhigh, max through output_config.effort; the default is medium.
Send system instructions separately from messages, provide max_tokens, and read text blocks from content. Public API requests support native streaming, image blocks, and client-side tool conversations.
Describe the objective, relevant context, constraints, and expected output. Keep source material distinct from instructions and ask for explicit evidence when reviewing documents.
FAQ
Use Anthropic Messages at /v1/messages with model anthropic/claude-opus-5-5. Include max_tokens and a messages array; pass system instructions in the top-level system field.
5-minute and 1-hour cache writes have separate rates. When the upstream provides both duration counts, billing uses those counts even if the reported total differs. If duration counts are unavailable, the reported write total is charged at the 5-minute rate.
Adaptive thinking is always enabled. Use output_config.effort with low, medium, high, xhigh, max; the default is medium.
Provide text and images through native Messages content blocks. The model returns text. The Playground accepts image attachments and retains the visible conversation for follow-up questions.
The context window is 1,000,000 tokens. Set max_tokens to a positive integer up to 128,000; this allowance includes thinking and final output.
Retain native content blocks, including signed thinking blocks, when continuing API tool-use conversations. Supply tool results in the native Messages format and keep prior conversation content intact.