AI NexusAI Nexus
Model releaseContains unconfirmed details

Anthropic releases Claude Sonnet 5.5: same list price, officially 30%+ faster

Event time 6 independent sourcesEditorial score 78/100Updated here

The short version

Anthropic released Claude Sonnet 5.5, the second model in the Claude 5.5 family, at the same list price as Sonnet 5: $2 per million input tokens and $10 per million output tokens. Anthropic says its output is more than 30% faster than Sonnet 5 and that, in its own testing, the cost per task is up to 30% lower (mainly because it needs fewer tokens to do the same work). The model is available on AWS, Google Cloud and Microsoft Azure.

What changed

The mid-tier Claude moves from Sonnet 5 to Sonnet 5.5 at an unchanged list price. In Anthropic's table, Terminal-Bench 4.0 rises from 10.3% to 70.6%, and it is the first Sonnet model to ship with cybersecurity safeguards and a fallback mechanism.

What it means for you

Developers who use Sonnet through the API can evaluate and switch at the same list price as Sonnet 5; billing in products such as Claude Code and GitHub Copilot follows each plan. "Up to 30% lower" is Anthropic's own testing, and actual spend depends on the task and thinking effort, so compare on your own tasks. Anthropic also states that Opus 5.5 remains clearly stronger for complex, open-ended work that needs sustained judgment.

Anthropic positions Sonnet 5.5 as a faster, lower-cost partner for Claude Opus 5.5: Opus handles complex work that needs careful judgment, while Sonnet 5.5 is good at well-scoped everyday tasks, fixing bugs, and producing documents, slides and spreadsheets. The same announcement says Claude Haiku 5.5 will join the 5.5 family in the coming weeks.

Migration note: Anthropic states that users who run Sonnet with thinking turned off need to switch to the new between_tools setting before migrating to Sonnet 5.5.

Third-party hands-on: In a test on launch day, Simon Willison found that generating an SVG at the "max" thinking effort made the model think for 128,000 tokens (about $1.28) and run out of tokens without producing anything, the same problem he had previously seen with Opus 5.5; the "xhigh" effort completed normally. This is a single hands-on test and not a general conclusion, but it is a reminder not to default to the highest effort.

Fact check

  • VerifiedSonnet 5.5 is priced at $2 per million input tokens, $10 per million output tokens and $0.20 for cache reads, the same as Sonnet 5Evidence
  • VerifiedAnthropic says output is more than 30% faster than Sonnet 5, and that in its own testing the cost per task is up to 30% lower than the previous generation because the same work usually takes far fewer tokensAnthropic's own testing; "up to", not a guaranteed value, and not an independent evaluation. Actual cost depends on the task and thinking effort.Evidence
  • VerifiedIn Anthropic's table, Terminal-Bench 4.0 is 70.6% versus 10.3% for Sonnet 5Vendor benchmark.Evidence
  • VerifiedSonnet 5.5 is the first Sonnet model to ship with cybersecurity safeguards and a fallback mechanism; higher-risk cybersecurity tasks noticeably fall back to Sonnet 5Evidence
  • VerifiedAnthropic says Opus 5.5 remains clearly stronger on complex, open-ended work that needs sustained judgment; Claude Haiku 5.5 will join the 5.5 family in the coming weeksEvidence
  • VerifiedUsers who run Sonnet with thinking turned off need to switch to the new between_tools setting before migrating to Sonnet 5.5Evidence
  • VerifiedAvailable on Amazon Web Services, Google Cloud and Microsoft AzureEvidence
  • VerifiedThe event time comes from Anthropic's official Claude Code v2.1.284 release (2026-09-28 18:02:03 UTC, which adds claude-sonnet-5-5); Anthropic's announcement page shows only the date, September 28, 2026TechCrunch's report time is 18:00 UTC, so the official announcement may be slightly earlier than this moment; this site does not backfill an earlier time on that basis.Evidence
  • VerifiedGitHub Copilot Pro, Pro+, Max, Business and Enterprise users can select Sonnet 5.5, counted toward usage-based billing at the vendor's list priceGitHub says it is rolling out gradually, so not seeing it yet is normal; administrators of Business and Enterprise can disable the model through model policies.Evidence
  • VerifiedSimon Willison's hands-on test: at the max thinking effort, the model thought for 128,000 tokens (about $1.28) and ran out of tokens without producing the SVG, while xhigh completed normallyA single hands-on test; it only shows that the author got this result and is not a general conclusion.Evidence
  • UnconfirmedSonnet 5.5 has become the model used by the free tier of claude.aiFrom Simon Willison's hands-on post; Anthropic's announcement page does not state this, and this site could not confirm it from an official source.Evidence

Coverage timeline

  1. CoverageTechCrunch
    Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner
  2. Primary sourceAnthropic (GitHub)
    Claude Code v2.1.284
  3. CoverageGitHub
    Claude Sonnet 5.5 in GitHub Copilot
  4. CoverageAWS
    Introducing Claude Sonnet 5.5 on AWS
  5. DiscussionSimon Willison
    Claude Sonnet 5.5
  6. Primary sourceAnthropic
    Introducing Claude Sonnet 5.5