Anthropic has introduced the discharge of Claude Haiku 4.5, which gives related coding efficiency to Claude Sonnet 4, however is twice as quick and one-third as costly.
This follows the discharge of Claude Sonnet 4.5 two weeks in the past, which Anthropic claimed as “the most effective coding mannequin on the planet” when it was introduced. Claude Haiku 4.5 outperforms Sonnet 4 on SWE-bench, however falls second behind Sonnet 4.5.
“Claude Sonnet 4.5, launched two weeks in the past, stays our frontier mannequin and the most effective coding mannequin on the planet. Claude Haiku 4.5 provides customers a brand new possibility for when they need near-frontier efficiency with a lot larger cost-efficiency,” Anthropic wrote in a publish.
The corporate defined that these two fashions can be utilized collectively to get the most effective of every, equivalent to utilizing Sonnet 4.5 to interrupt down a fancy downside into a number of steps after which having a number of Haiku 4.5 cases working collectively to execute these subtasks in parallel.
In response to Anthropic, Haiku 4.5 excels at real-time, low-latency duties, equivalent to chat assistants, customer support brokers, and pair programming.
It would price $1 per million enter tokens and $5 per million output tokens. For comparability, Claude Sonnet 4 and 4.5 price $3 per million enter tokens and $15 per million output tokens.
Claude Haiku 4.5 is accessible now in Claude Code and different Anthropic apps, in addition to by means of the API, Amazon Bedrock, and Google Cloud’s Vertex AI.
