Claude Haiku 5.5
Anthropic · Lançado 10/2026
Claude Haiku 5.5 is a conditional pick: Anthropic's fastest and cheapest current Claude at $0.10 input and $0.50 output per 1M tokens, with a 1M context and adjustable effort. Prompts over 100K tokens cost five times as much, and every benchmark is Anthropic's own. Right for high-volume classification, extraction, routing and subagent work. Use Sonnet 5.5 for complex agentic coding, where Anthropic's own Terminal-Bench 4.0 numbers put it well ahead.
É adequada para ti?
Bom para
- High-volume, latency-sensitive work such as classification, extraction, routing and subagent tasks, which is the use Anthropic names for it
- Cost at scale: $0.10 input and $0.50 output per 1M tokens up to 100K-token prompts, cache reads $0.01, Batch API at 50% off
- Long inputs: 1M-token context and 128K output, with 300K output on the Batches API behind a beta header
- Multi-cloud buyers: the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS all offer Haiku 5.5
Não recomendado para
- Cost-sensitive prompts over 100K tokens: input rises to $0.50 and output to $2.50 per 1M, five times Haiku 5.5's own base rate, and cache reads count toward the threshold
- Complex agentic coding: Anthropic's own table has Haiku 5.5 at 39.2% on Terminal-Bench 4.0 against Sonnet 5.5 at 70.6%
- Code that sets temperature, top_p or top_k: any non-default value returns a 400 error
- Penetration testing: Anthropic says the safeguards still block it
- Decisions that need independently reproduced scores: every published number is Anthropic's
Desempenho por tarefa
Classification and extraction
Anthropic states this as the model's purpose; vendor-stated, no benchmark for it was read, and untested by us
Agentic coding
Terminal-Bench 4.0 39.2% and FrontierCode 1.1 46.4% in Anthropic's table, both below Sonnet 5.5; vendor-reported, untested by us
Computer use
OSWorld 2.1 offline subset 72.4% in Anthropic's table, against 83.9% for Sonnet 5.5; vendor-reported, untested by us
Preços
Entrada
$0.10 / 1M tokens
Saída
$0.50 / 1M tokens
Contexto
1M ctx; prompts over 100K cost $0.50/$2.50
Testes de desempenho
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
Ainda não há verificações
Ainda não registámos nenhuma verificação para esta entrada. Assim que uma for executada, o seu histórico aparece aqui.