Pricing verified October 6, 2026
Claude Sonnet 5.5 Pricing & Specs
Anthropic’s mid-tier workhorse since September 28, 2026: $2/$10 per million tokens, 1M context, officially 30% faster than Sonnet 5 at up to 30% lower cost for most work.
Sonnet 5.5 at a glance
Claude Sonnet 5.5 (claude-sonnet-5-5) launched September 28, 2026 at $2 per million input tokens and $10 per million output — the same list price as Sonnet 5. Cache writes cost $2.50 (5-minute) or $4 (1-hour) per million tokens; cache hits are $0.20; the Batch API halves everything to $1/$5. It has a 1M-token context window, 128K max output and a June 2026 knowledge cutoff, and Anthropic says it runs 30% faster and costs up to 30% less than Sonnet 5 for most work. Sonnet 5 moved to Additional models the same day, and Anthropic documents five ways code written for Sonnet 5 can break. All figures verified against Anthropic’s pricing page as of October 6, 2026.
Claude Sonnet 5.5 specs & price table
Every number from Anthropic’s pricing page and models overview, verified October 6, 2026.
| Spec | claude_sonnet_5_5_page.table.col_1 |
|---|---|
| Released | 2026-09-28 |
| API model ID | claude-sonnet-5-5 |
| Input (per 1M tokens) | $2.00 |
| Output (per 1M tokens) | $10.00 |
| Cache hit | $0.20 |
| Cache write (5-minute) | $2.50 |
| Cache write (1-hour) | $4.00 |
| Batch (in / out) | $1.00 / $5.00 |
| Context window | 1M |
| Max output | 128K |
| Knowledge cutoff | June 2026 |
| Thinking | Adaptive (not always-on) |
| Default effort | effort parameter, default high |
| Availability | Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, Microsoft Foundry |
| Data retention | Not stated on the pricing page |
All rates above are read from our MODEL_REGISTRY and verified against Anthropic’s pricing page on 2026-10-06 (USD per million tokens).
Claude Sonnet 5.5 context window & specs
The context numbers people actually search for, verified against Anthropic’s models overview on October 6, 2026.
1M-token context window
Sonnet 5.5 accepts up to 1,000,000 tokens of context — the same window as Opus 5.5 and Fable 5.1, and five times the 200K window on Haiku 4.5. Long documents, large codebases and whole-repo agent sessions fit without chunking.
128K max output tokens
Up to 128,000 output tokens per response — enough for large single-file refactors and long agentic trajectories.
Knowledge cutoff: June 2026
Training data runs through June 2026 — roughly four months fresher than the models it replaced in the main pricing table. For anything newer, pair it with web search or a retrieval step.
Adaptive thinking, not always-on
Sonnet 5.5 thinks adaptively but, unlike Opus 5.5 and Fable 5.1, thinking is not always-on — you can still turn it off (using the new between_tools syntax, see the breaking changes below). Default effort is high.
Taken together: a 1M context window, 128K max output and a June 2026 cutoff put Sonnet 5.5 on the same long-context footing as the flagship tier, at one fifth of Fable 5.1’s sticker price (as of 2026-10-06).
Sonnet 5.5 vs Sonnet 5: same price, faster model
List prices did not move. What changed is speed and efficiency — plus five ways code written for Sonnet 5 can break.
| Metric | Sonnet 5.5 | Sonnet 5 |
|---|---|---|
| Input (per 1M tokens) | $2.00 | $2.00 |
| Output (per 1M tokens) | $10.00 | $10.00 |
| Cache hit | $0.20 | $0.20 |
| Cache write (5-minute) | $2.50 | $2.50 |
| Cache write (1-hour) | $4.00 | $4.00 |
| Batch (in / out) | $1.00 / $5.00 | $1.00 / $5.00 |
| Context window | 1M | 1M |
| Knowledge cutoff | June 2026 | Not confirmed |
Every sticker rate is identical: $2 input, $10 output, $0.20 cache hits (0.1x input — the same 0.1x multiplier Sonnet 5 uses), $2.50/$4 cache writes and $1/$5 Batch. Anthropic’s own claim is that Sonnet 5.5 “runs 30% faster and costs up to 30% less for most work” — with list prices unchanged, that saving comes from efficiency (fewer tokens, faster turns), not from a price cut. Sonnet 5 remains available in Additional models at $2/$10, so nothing forces a migration; but Sonnet 4.5 retires on November 30, 2026, and that workload flows to 5.5 — where the five breaking changes below do apply to Sonnet 5 code.
Five breaking changes for Sonnet 5 code
From Anthropic’s release note of September 28, 2026 — check each one before pointing an existing Sonnet 5 pipeline at 5.5.
Turning thinking off has new syntax
To disable up-front thinking, send a thinking parameter of type between_tools instead of disabled — and only at high effort or below. The old disabled value no longer behaves as before.
Forced tool use returns 400
tool_choice types any and tool now return a 400 error. If you were forcing a tool call, switch to auto with strict tool use.
Thinking blocks are tied to model and conversation
Passing thinking blocks across models or conversations no longer works — blocks produced in a Sonnet 5 run cannot be replayed on 5.5.
Old computer-use tool rejected on some platforms
On the Claude API and Google Cloud, the earlier computer_20251124 computer-use tool is not accepted. The release note names no replacement for those platforms — check the current toolset docs before migrating.
Advisor tool rejects older models as advisors
The advisor tool no longer accepts Claude Opus 4.8, Opus 4.7 or Sonnet 5 as advisors.
Sonnet 5.5 vs Opus 5.5: the cheaper sibling
Two models, one main pricing table: Anthropic positions Sonnet 5.5 as a faster, lower-cost complement to Opus 5.5.
| Metric | Sonnet 5.5 | Opus 5.5 |
|---|---|---|
| Input (per 1M tokens) | $2.00 | $4.00 |
| Output (per 1M tokens) | $10.00 | $20.00 |
| Cache hit | $0.20 | $0.20 |
| Cache write (5-minute) | $2.50 | $5.00 |
| Cache write (1-hour) | $4.00 | $8.00 |
| Batch (in / out) | $1.00 / $5.00 | $2.00 / $10.00 |
Sonnet 5.5 costs 50% less than Opus 5.5 on input and 50% less on output ($2/$10 vs $4/$20), with identical 1M context windows and 128K max output. The behavioral difference is thinking: Opus 5.5 is always-on adaptive (it cannot be disabled), while Sonnet 5.5 keeps an on/off switch. As of 2026-10-06, Sonnet 5.5 is not in the Fast mode preview — that table still lists only Opus 5.5 ($8/$40, exactly 2x its base rate) and the older Opus models.
Where Sonnet 5.5 is available
Platform availability and model IDs, verified October 6, 2026.
Claude API
Model ID claude-sonnet-5-5, live since September 28, 2026. Rates: $2 in / $10 out per million tokens, cache hits $0.20, cache writes $2.50 (5-minute) / $4 (1-hour), Batch $1/$5.
Cloud platforms
Available on Amazon Bedrock (anthropic.claude-sonnet-5-5), Claude Platform on AWS, Google Cloud (claude-sonnet-5-5@…) and Microsoft Foundry (same model ID as the Claude API).
No Fast mode tier
As of October 6, 2026, Sonnet 5.5 is not in Anthropic’s Fast mode research preview — the Fast mode table still lists only Opus 5.5 ($8/$40) and Opus 5 / Opus 4.8 ($10/$50).
Migrating from Sonnet 4.5 (or Sonnet 5)
Anthropic deprecated Sonnet 4.5 on September 30, 2026, with API retirement on November 30, 2026, and recommends migrating to Sonnet 5.5. Sonnet 5 itself is not deprecated — it sits in Additional models at unchanged prices — but audit your code for the five breaking changes above before switching the model ID.
This page was last reviewed 2026-10-06. Anthropic moved Sonnet 5 to Additional models on September 28, 2026 and announced Sonnet 4.5’s retirement (November 30, 2026) two days later — the mid-tier line is moving fast, so re-check before you commit new code. Anthropic’s pricing page remains the source of record.
Estimate your Sonnet 5.5 bill
Run your workload through the cost calculator — input, output, cache-hit rate, effort level and Batch routing — to see what $2/$10 means for your monthly spend.