Claude Sonnet 5.5

What Anthropic says. The choosing guide's selection matrix lists Sonnet 5.5 for "speed and capability for everyday coding, agent, and enterprise workloads", with code generation, data analysis, content creation, visual understanding and agentic tool use as examples. The guide says to start at the default high effort and adjust based on your evals. The deprecations page names claude-sonnet-5-5 as the recommended replacement for Claude Sonnet 4.5, which retires on 30 November 2026, and for several older Sonnet models.

Our editorial judgment. Sonnet 5.5 looks like the natural production tier for teams that have prototyped on Opus 5.5 and now need to control spend. The option to turn off up-front thinking is useful for interactive tools where response time matters more than depth. For short, repetitive jobs such as tagging or routing, Claude Haiku 5.5 is listed at a fraction of the price and is worth testing first. If you are picking a packaged coding assistant rather than an API model, see our AI coding tools guide.

• Five breaking changes from Claude Sonnet 5. The model page lists: turning off up-front thinking now uses between_tools; forced tool use returns an error; thinking blocks are tied to the model and conversation; on the Claude API and Google Cloud the earlier computer_20251124 tool is not accepted; and the advisor tool rejects Claude Opus 4.8, Claude Opus 4.7, Claude Sonnet 5 and Claude Haiku 5.5 as advisors.
• between_tools has an effort ceiling. The lowest thinking setting works only at high effort or below.
• Quiet streams between tool calls. Text between tool calls comes back in thinking blocks. Apps that stream that text go quiet until they set a display value or switch to between_tools.
• Sampling parameters. Setting temperature, top_p or top_k to a non-default value returns a 400 error.
• No fast mode. The pricing page lists fast mode only for Opus 5.5, Opus 5 and Opus 4.8.
• US-only inference costs more. Setting inference_geo to "us" applies a 1.1x multiplier to all token prices for Claude 4.6 and later models.
• Tokenizer. Anthropic says Claude 4.7 and later models produce about 30% more tokens for the same text than earlier models.

All four current models share a 1M token context window, 128K max output, text and image input and a June 2026 reliable knowledge cutoff, according to the models overview.

Against its predecessor, the pricing page lists Claude Sonnet 5 at the same USD 2 input and USD 10 output, but with cache hits at USD 0.20 per 1M tokens, double the Sonnet 5.5 rate. Claude Sonnet 4.6 and 4.5 are listed at USD 3 input and USD 15 output.

1. Take 20 to 50 tasks with known good outputs, including some long tool-using runs.
2. Run them on Sonnet 5.5 at the default high effort and on Opus 5.5 at medium, with identical prompts and tools, scoring each against your own rubric.
3. For interactive features, rerun the set with between_tools and record latency alongside pass rate.
4. Multiply logged input, cached and output tokens by list prices to get cost per completed task, and move a task type to Sonnet 5.5 only where its pass rate holds up.

We read Anthropic's models overview, Sonnet 5.5 model page, pricing page, choosing a model guide and deprecations page on 15 October 2026. We have not run Claude Sonnet 5.5, and nothing here is a test result or ranking. Check the vendor pages before committing budget. See our ranking method for how HumanRank scores tools and the model directory for other profiles.

What is Claude Sonnet 5.5? It is Anthropic's mid-tier model, positioned as a combination of speed and intelligence for everyday coding, agent and enterprise workloads.

How much does Claude Sonnet 5.5 cost? It costs USD 2 per 1M input tokens and USD 10 per 1M output tokens at list price on 15 October 2026, with cache hits at USD 0.10 per 1M tokens.

What is the Claude Sonnet 5.5 context window? It is 1M tokens, with up to 128K output tokens, or 300K on the Message Batches API with the output-300k-2026-03-24 beta header.

Can I turn off thinking on Claude Sonnet 5.5? Partly, the between_tools setting turns off up-front thinking, and it works at high effort or below.

Does Claude Sonnet 5.5 replace Claude Sonnet 4.5? Yes, Anthropic names it as the recommended replacement for Claude Sonnet 4.5, which is deprecated and retires on 30 November 2026.

Is Sonnet 5.5 cheaper than Opus 5.5? Yes, its list token prices are half of Opus 5.5's, though whether it costs less per finished task depends on your own test results.