# tools.strict/search: strict tool use (models: https://platform.claude.com/docs/en/build-with-claude/structured-outputs) and the tool search tool on the Messages API: https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool#model-compatibility # tools.choice: auto/any/tool/none; Opus 5.5, Sonnet 5.5 and Fable 5.1 reject forced tool use (any/tool): https://platform.claude.com/docs/en/agents-and-tools/tool-use/define-tools#forcing-tool-use # Web search: $10 per 1,000 searches; each search counts once, however many results it returns: https://platform.claude.com/docs/en/about-claude/pricing#web-search-tool # Token-only multipliers (object factor, so cost.web_search is unscaled): US data residency (https://platform.claude.com/docs/en/manage-claude/data-residency#pricing). # APIs: Messages, and Chat Completions through the OpenAI SDK compatibility layer (https://platform.claude.com/docs/en/api/openai-sdk) # Per-API tools: the OpenAI SDK compatibility layer ignores `strict` and takes only function tools, so strict and search sit on [api.messages] only; tool_choice works on both APIs and stays top-level (https://platform.claude.com/docs/en/api/openai-sdk). Live 2026-10-26 on Chat Completions (claude-sonnet-4-6): a strict tool whose schema strict mode rejects was accepted, and a tool_search_tool_regex tool returned 400 "tools.0.type: Input should be 'function'". schema = 2 name = "Claude Sonnet 4.6" description = "Claude workhorse for coding agents, careful analysis, and production cost control" family = "claude-sonnet" open_weights = false knowledge = "2025-08-31" release_date = "2026-02-17" last_updated = "2026-03-13" [modalities] input = ["text", "image/jpeg", "image/png", "image/gif", "image/webp", "application/pdf"] output = ["text"] [capabilities] tools = { supported = true, choice = ["auto", "none", "required", "tool"] } reasoning = { supported = true, effort = ["low", "medium", "high", "max"], budget = { min = 1_024 } } structured_output = true temperature = true cache = { explicit = { ttl = ["5m", "1h"], max_breakpoints = 4, targets = ["tools", "system", "messages"], min_tokens = 1_024 } } [limit] context = 1_000_000 output = 128_000 [api.messages] capabilities = { tools = { strict = true, search = true } } [api.chat-completions] base_url = "https://api.anthropic.com/v1" sdk = { aisdk = "@ai-sdk/openai-compatible", opencodeai = "@opencode/ai/providers/openai-compatible" } [cost.text.input] price = 3 cache = { read = 0.3, write = { "5m" = 3.75, "1h" = 6 } } [cost.text.output] price = 15 [cost.web_search] price = 10 [[cost.multipliers]] when = { region = { equals = "us" } } factor = { text = 1.1 }