Claude Sonnet 5.5 API: pricing and capabilities

Anthropic released Claude Sonnet 5.5 on September 28, 2026 with the API model ID claude-sonnet-5-5. This TTAPI guide brings its published model specifications and token rates into the existing Claude family layout. Anthropic positions Sonnet 5.5 between its lower-cost, everyday work and the harder open-ended tasks for which it recommends Opus 5.5.

The launch describes faster answers, fewer tokens per completed task, better bug fixing, and stronger document, slide, and spreadsheet work than Sonnet 5. Those are provider claims about its tests. A useful upgrade decision starts with a small set of your own tasks and checks the finished result, latency, and total token usage.

The Claude Platform lists a 1M-token context window and a 128K-token maximum output for Sonnet 5.5. It accepts text and images as input and produces text. Its documented reliable knowledge cutoff is June 2026. A long context window is capacity, not a guarantee that every fact in a long document will be found or cited correctly.

For large repositories or document collections, start with a scoped bundle of relevant files, require the answer to identify its evidence, and test retrieval quality as the bundle grows. Reserve enough output budget for the requested deliverable; a 128K maximum does not mean every request should ask for 128K output tokens.

Anthropic reports that Sonnet 5.5 generates output more than 30% faster than Sonnet 5. In its Terminal-Bench 4.0 agentic coding evaluation, the launch reports 70.6% for Sonnet 5.5 versus 10.3% for Sonnet 5. These scores come from Anthropic's published setup; they are not measurements of TTAPI latency or success rate.

For a coding agent, compare both versions on the same failing test or issue, with the same tools and review rules. Record whether the patch passes, how many unrelated files changed, how many tool calls it took, and the cost of an accepted result. Faster token output can help iteration, but a failed patch still needs human review and another run.

Anthropic lists Sonnet 5.5 at $2 per million input tokens, $10 per million output tokens, and $0.2 per million cache-read tokens. The matching Opus 5.5 rates are $4, $20, and $0.2. Sonnet therefore halves the standard input and output token price versus Opus, while the cache-read rate is the same.

The TTAPI Claude price section follows official model pricing and shows Sonnet 5.5 in the same family table. Anthropic separately lists a $2.50 five-minute cache-write rate and a $4 one-hour cache-write rate per million tokens. These write rates are not represented by the TTAPI table's cache-read column. TTAPI's published guide does not yet list Sonnet 5.5, so check live model availability and billing before using these displayed rates as a production quote.

Anthropic describes Sonnet 5.5 as a fit for well-scoped everyday tasks, bug fixes, and polished documents; it describes Opus 5.5 as stronger on complex, open-ended work that needs sustained judgment. This is a useful starting hypothesis, not a rule for every application. A difficult production debugging task may still justify Opus, while a repeatable extraction or drafting job may favor Sonnet.

Price per million tokens is only one part of the decision. Compare cost per accepted answer after retries, tool calls, and reviewer corrections. If Sonnet uses fewer tokens than an older Sonnet model on your workload, the cost can fall even when the unit price stays unchanged. Anthropic reports savings of up to 30% per task versus Sonnet 5 in its tests; measure that claim on your own prompts before forecasting spend.

The Claude Platform's Sonnet 5.5 migration notes include changes to adaptive thinking and tool behavior. Turning off up-front thinking now uses the between_tools setting. Forced tool use can return an error, and text between tool calls may arrive in thinking blocks unless an appropriate display setting is used. Applications that parse streamed responses or depend on forced tool selection should test those paths before changing the model ID.

Those are Anthropic Platform behaviors. TTAPI's Messages route documents the common model, max_tokens, and messages fields; it does not establish that every provider-specific thinking or display control is exposed. Keep an existing working model as a fallback and validate response parsing, tool handoffs, and billing in a test account before rollout.

The request below uses TTAPI's documented Claude Messages shape with the new Anthropic model ID. Treat it as an integration example: Anthropic confirms the ID, but TTAPI's public supported-model list has not yet named Sonnet 5.5. Check the model list in your TTAPI account and confirm that the request succeeds before depending on this route.

Start with one bug-fix or document-review task that has a clear acceptance check. Save response text, usage, time to completion, and any errors. Run the same task through Opus 5.5 if you need a quality and cost comparison. Once the result is acceptable and billing matches the current account price, move the model ID into your server-side configuration.

curl --request POST \
  --url 'https://api.ttapi.io/v1/messages' \
  --header "x-api-key: $TTAPI_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "claude-sonnet-5-5",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": "Review this bug report and patch. Identify one likely regression, suggest the smallest fix, and list two tests to run."
    }
  ]
}'

• Confirm claude-sonnet-5-5 appears in the TTAPI account's available models.
• Keep TTAPI_KEY on the server and send it as x-api-key.
• Test usage, output parsing, and final billed rate with a small request.