Anthropic has introduced the discharge of Claude Haiku 4.5, which supplies comparable 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 “one of the best 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 one of the best coding mannequin on the planet. Claude Haiku 4.5 offers customers a brand new choice for when they need near-frontier efficiency with a lot better cost-efficiency,” Anthropic wrote in a submit.
The corporate defined that these two fashions can be utilized collectively to get one of the best of every, reminiscent of utilizing Sonnet 4.5 to interrupt down a fancy drawback 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, reminiscent of chat assistants, customer support brokers, and pair programming.
It’s going to value $1 per million enter tokens and $5 per million output tokens. For comparability, Claude Sonnet 4 and 4.5 value $3 per million enter tokens and $15 per million output tokens.
Claude Haiku 4.5 is offered now in Claude Code and different Anthropic apps, in addition to by means of the API, Amazon Bedrock, and Google Cloud’s Vertex AI.
