Anthropic unveiled Claude Haiku 4.5, a compact language model designed for fast coding tasks. The company says it can deliver performance comparable to its larger frontier model, Claude Sonnet 4, but at a fraction of the cost and with higher speed.
Claude Haiku 4.5 sits in the Haiku line alongside Sonnet (medium) and Opus (large). Through distillation, Anthropic compresses larger models into smaller ones that still perform well on coding and related tasks, albeit with some knowledge retention trade-offs.
In benchmarks shared by Anthropic, Haiku 4.5 reportedly matches five-month-old coding capabilities of Sonnet 4 and even approaches GPT-5 in some coding tasks, while running faster and cheaper. Independent testers will be needed to confirm these results, but the claims highlight the potential of distillation to shave costs without sacrificing critical performance.
For users needing quick, concept-to-code assistance, Haiku 4.5 can be attractive due to speed and lower cost. For broader, deeper analysis on foreign policy or history, Sonnet or Opus may still be preferable despite potential errors in any AI model.
Claude Haiku 4.5 is available now to all Claude app, web, and API users. In pricing terms, Anthropic positions Haiku 4.5 at about $1 per million input tokens and $5 per million output tokens, with Sonnet 4.5 at $3 input / $15 output, and Opus 4.1 at $15 input / $75 output, making Haiku the most economical option in the Claude lineup for coding-oriented workloads.
