Claude Sonnet 5.5 logo

Claude Sonnet 5.5

Anthropic · Lançado 09/2026

Condicional

Conditional. Claude Sonnet 5.5 is Anthropic's default Sonnet model on the Anthropic API, on the same rate card as Sonnet 5, and Anthropic says it runs faster and costs less per task. Every benchmark is Anthropic's own, and five breaking changes hit code written for Sonnet 5. Right for teams ready to re-test agentic coding and knowledge work; hold off if you rely on forced tool choice or temperature control.

É adequada para ti?

Bom para

  • Agentic coding and command-line work; Anthropic reports 70.6% on Terminal-Bench 4.0, unreproduced by us
  • Large-context work: 1M-token context and 128K output, 300K output on the Batches API with a beta header
  • Cost control at scale: $2 input and $10 output per 1M tokens, batch at half price and cache reads at $0.20
  • Multi-cloud deployment: the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS
  • Claude Code users: Sonnet 5.5 (claude-sonnet-5-5) was added in Claude Code 2.1.284 on Sep 28, 2026

Não recomendado para

  • Code that forces tool use: forced tool use returns an error
  • Apps that set temperature, top_p or top_k: non-default values return a 400 error
  • Interfaces that stream text between tool calls without setting a display value or between_tools: they go quiet
  • Higher-risk cybersecurity tasks, which Anthropic says visibly fall back to Sonnet 5
  • Anything beyond text and image input with text output

Desempenho por tarefa

Agentic coding

Very Good

Anthropic-reported Terminal-Bench 4.0 and CursorBench 4.0 results; vendor-reported, untested by us

Knowledge work

Very Good

Anthropic-reported GDPval-AA v2.1 of 1844; vendor-reported, untested by us

Cybersecurity

Fair

Higher-risk cyber tasks visibly fall back to Sonnet 5 by design

Preços

Entrada

$2 / 1M tokens

Saída

$10 / 1M tokens

Contexto

1M ctx; batch $1/$5; cache read $0.20

Ver preços completos

Testes de desempenho

TestePontuaçãoFonte
Terminal-Bench 4.0 (Anthropic-reported)70.6% Fonte
CursorBench 4.0 (Anthropic-reported)55.5% Fonte
GDPval-AA v2.1 (Anthropic-reported)1844 Fonte
Humanity's Last Exam, with tools (Anthropic-reported)64.5% Fonte

Ainda sem alterações de veredicto

O relógio começa a contar no primeiro dia: as alterações aparecem aqui à medida que o nosso veredicto evolui.

Registo de verificação

  • Preços· Sem alterações

    Agente automatizado

  • Preços· Sem alterações

    Agente automatizado

Como avaliamos