Claude Opus 5.5
Anthropic · Lançado 09/2026
Claude Opus 5.5 is the model Anthropic now recommends starting with for most workloads, built for long-running agentic coding and knowledge work at rates 20% below Claude Opus 5. Conditional because every benchmark is Anthropic's own and migration is not drop-in: thinking cannot be turned off, forced tool use returns an error, and most cybersecurity tasks are rerouted to Opus 4.8. Right for Claude Opus 5 teams that can absorb those breaking changes.
É adequada para ti?
Bom para
- Long-running agentic coding and knowledge work, the use Anthropic built it for, with a 1M-token context and 128K output (300K on the Batch API in beta)
- Cutting cost from Claude Opus 5: $4/$20 per 1M tokens against $5/$25, cache reads $0.20 against $0.50, batch $2/$10
- Cache-heavy agent loops: cache hits cost 5% of the input price, half the standard 10% ratio
- Latency-sensitive work on the Claude API: fast mode (research preview) runs $8/$40 per 1M tokens, below Opus 5's $10/$50 fast rate
Não recomendado para
- Drop-in upgrades from Opus 5: thinking cannot be disabled, forced tool use returns an error, thinking blocks are tied to the model and conversation, and computer_20251124 is refused on the Claude API and Google Cloud
- Apps that stream progress text between tool calls: that text now returns in thinking blocks that are empty at the default display setting
- Cybersecurity work: Anthropic says most cyber tasks are rerouted to Opus 4.8, and its Cyber Verification Program has not yet been expanded to include Opus 5.5
- Decisions that need independent benchmarks: every published score is Anthropic's own
Preços
Entrada
$4 / 1M tokens
Saída
$20 / 1M tokens
Contexto
1M ctx; 128K out; cache $0.20; fast $8/$40
Testes de desempenho
| Teste | Pontuação | Fonte |
|---|---|---|
| Terminal-Bench 4.0 (Anthropic-reported, max effort) | 66.4% | Fonte |
| FrontierCode v1.1 Main (Anthropic-reported, max effort) | 54.4% | Fonte |
| CursorBench 4.0 (Anthropic-reported, max effort) | 57.8% | Fonte |
| GDPval-AA v2.1 (Anthropic-reported, max effort) | 1846 Elo | Fonte |
| Humanity's Last Exam, with tools (Anthropic-reported, max effort) | 67.7% | 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.
Fontes
Registo de verificação
- Preços· Sem alterações
Agente automatizado
- Preços· Sem alterações
Agente automatizado
- Preços· Sem alterações
Agente automatizado
- Perfil· Sem alterações
Agente automatizado
- Preços· Sem alterações
Agente automatizado