Claude Sonnet 5.5 API
NewLLMThe Claude Sonnet 5.5 API is Anthropic's newest Sonnet route: it keeps Sonnet 5's $2 / $10 rate card and 1M context, and adds always-on adaptive thinking to the balanced vision-and-tools mix.
- Input · USD / 1M tokens$0.56 USDOfficial price
$2 USD - Output · USD / 1M tokens$2.8 USDOfficial price
$10 USD - Savings72%
- Context1M
- View tool pricing and billing rules
What are we making today?
Bring an idea, a question, or a first draft.
01
What this model is for
The Claude Sonnet 5.5 API is Anthropic's newest Sonnet route: it keeps Sonnet 5's $2 / $10 rate card and 1M context, and adds always-on adaptive thinking to the balanced vision-and-tools mix.
Claude Sonnet 5.5 shipped on September 28, 2026, six days after Opus 5.5, and fills the middle of Anthropic's Claude 5.5 range. It matches the Opus release on capacity: a 1M-token context window, 128K standard output, text and image input, tool use. Two catalog details set it apart from Sonnet 5: adaptive thinking is now listed, while the token rates stay as they were. Compared with Opus 5.5 it gives up depth for a lower input and output rate and keeps the same cache-read rate.
Sonnet 5.5 suits jobs where a machine can judge the result cheaply: a scoped feature with a test suite, a reproduced bug, a focused review, a draft assembled from supplied sources. Its API default effort is high, one step above Opus 5.5, and it adds a between_tools thinking mode for turns that should start answering without up-front reasoning. Escalate to Opus 5.5 when repeated attempts keep misreading the cause rather than the fix.
Best-fit coding workloads
Feature work behind tests
State the change, point at the existing test file and ask for a patch plus new cases that fail before it. A bounded task with a runnable suite gives the model a cheap pass or fail gate, so a weak patch is rejected by the suite instead of by a reviewer.
Second opinion on a pull request
Send the diff, the tests that cover it and the issue it closes, then ask for defects ranked by severity with the line that shows each one. Keep the turn read-only. Running the same prompt on Opus 5.5 shows whether the extra tier finds anything the first pass missed.
Draft documents from supplied sources
Supply the report or the PDFs plus an outline, and require a pointer to the source page behind every claim. The window holds a long document in one request, and the same conversation can return slide text or spreadsheet formulas for the sections that already pass review.
02
Hands-on notes
Checked on 2026-10-27 through the OmniaKey API with one standard coding prompt (median of 3 runs) and three capability probes.
- Output tokens for our standard coding prompt, reasoning included
- 609
- Cost of that output at current prices
- $0.00171
- JSON mode
- Passed
- Tool calling
- Passed
- Image input
- Passed
03
Technical specification
Catalog facts and integration details verified for this route.
- Model ID
- claude-sonnet-5-5
- Context
- 1M
- Inputs and tools
- Vision · Tool use · Adaptive thinking
- API endpoint
- /v1/chat/completions
- Cache hit / 1M
- $0.056 USD
- Coding clients
- Claude Code · Cursor · Cline · Aider
Limits to plan around
- Non-default temperature, top_p and top_k values are rejected, so a client that hardcodes old sampling defaults must drop them before the first call. Check what your SDK adds on its own.
- Thinking blocks stay tied to the model and the conversation. This model can read Sonnet 5 blocks, but Opus 5 and Opus 5.5 cannot read the blocks it writes, so a quiet fallback between models loses them.
- Progress text between tool calls may be delivered inside thinking blocks whose default display is empty, which makes a text-only client look stalled while the answer is still being assembled.
- The 1M window is capacity, not guaranteed recall, and max_tokens has to cover billed thinking as well as the visible answer on long generations.
04
Compare nearby models
Compare workload fit and current listed token rates before choosing a route.
| Model | Best for | Context | Input | Output |
|---|---|---|---|---|
| Claude Sonnet 5.5This page claude-sonnet-5-5 | The daily choice for scoped coding and document work that a test, schema or reviewer can judge. | 1M | $0.56 USD Official price $2 USD | $2.8 USD Official price $10 USD |
| Claude Opus 5.5 claude-opus-5-5 | Ambiguous diagnosis, cross-service failures and decisions where a second pass costs more than the tokens. | 1M | $1.12 USD Official price $4 USD | $5.6 USD Official price $20 USD |
| Claude Sonnet 5 claude-sonnet-5 | Legacy Sonnet 5 pins that still need their prompts and parsers replayed before moving to 5.5. | 1M | $0.56 USD Official price $2 USD | $2.8 USD Official price $10 USD |
| Claude Haiku 5.5 claude-haiku-5-5 | High-volume classification, extraction and small subagent steps with a clear, inexpensive check. | 1M | $0.08 USD Official price $0.1 USD | $0.4 USD Official price $0.5 USD |
05
SDK quickstarts
Use one base URL and model ID from cURL, Python, or TypeScript.
curl https://api.omniakey.com/v1/chat/completions \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5-5",
"messages": [
{"role": "user", "content": "Review this pull request"}
],
"stream": true
}'06
Estimate token cost
Enter expected input and output tokens to compare current listed rates.
- OmniaKey estimate
- $0.112
- Official-rate estimate
- $0.40
- Estimated savings
- $0.288
Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.
07
Related coding guides
Deeper selection, setup, and troubleshooting guidance for this model family.
Claude Sonnet 5.5 review
A faster daily model with a more important API migration than the version number suggests.
Read guideOpus 5.5 vs Sonnet 5.5
Choose by results and total cost.
Read guideBest model for Claude Code
Start daily coding with Sonnet 5.5 at $2 per million input tokens, move complex multi-file work to Opus 5.5 at $4, keep Fable 5.1 for the hardest long sessions, and hand quick edits and subagents to Haiku 5.5.
Read guideClaude Pricing 2026: Every Plan, Usage Limit and Model Included
Every Claude plan, its price and its usage limit in one place, plus the API rates and the plan that actually includes Fable.
Read guide08
Sources and verification
Model facts and links were checked on 2026-10-27. Provider documentation remains the authority for upstream behavior.
Frequently asked questions
How much does Claude Sonnet 5.5 cost on OmniaKey?
OmniaKey charges $0.56 USD per 1M input tokens and $2.8 USD per 1M output tokens, in USD. Official prices are $2 USD and $10 USD; you save 72%.
What is the context window for Claude Sonnet 5.5?
The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.
When was Claude Sonnet 5.5 released?
Anthropic released Claude Sonnet 5.5 on September 28, 2026. Its Claude API model id is claude-sonnet-5-5, and the OmniaKey catalog uses the same id. On Anthropic's API the sonnet alias points at this model, but a pinned id is safer in evaluations and in production, because aliases can move.
Is Claude Sonnet 5.5 better than Claude Opus 5.5?
There is no single answer. Sonnet 5.5 is the cheaper starting point for work with a reliable check, and Anthropic reports results close to Opus 5.5 on several coding and knowledge-work evaluations. Opus 5.5 keeps thinking always on and is positioned for harder, open-ended judgment. Compare finished tasks, retries and review minutes on your own repository.
How much does Claude Sonnet 5.5 cost?
Billing is per token: uncached input, output that includes billed thinking, cache reads, and cache writes at five-minute or one-hour lifetimes. This model has one rate card across the whole 1M window, with no step for long prompts, and cache reads are charged at their own rate. Current numbers are listed in the price table on the model listing.
Can I turn thinking off on Claude Sonnet 5.5?
You cannot send the old disabled setting; it returns an error, as does a manual budget_tokens value. Instead leave thinking unset for adaptive behaviour, or set thinking to between_tools, which skips up-front thinking at high effort and below. Effort is controlled separately with output_config.effort.
Can I swap Sonnet 5.5 in for Sonnet 5 without changing code?
Usually not. A Sonnet 5 request that disables thinking, forces a tool call, or sends non-default sampling values fails against this model. Parsers that read the response by position rather than by block type can also break. Replay a few real requests and check tool loops before switching a running agent over.
How do I call Claude Sonnet 5.5 through OmniaKey?
Send POST /v1/chat/completions to https://api.omniakey.com/v1 with claude-sonnet-5-5 as the model, using any OmniaKey key. In Claude Code, set ANTHROPIC_BASE_URL to https://api.omniakey.com and ANTHROPIC_AUTH_TOKEN to the key, then pick the model id; version 2.1.284 or later is needed for this model to appear in the picker.