Anthropic’s Claude Haiku 4.5 matches May’s frontier model at a fraction of the cost

And speaking of costs, Haiku 4.5 is included for subscribers of Claude’s web and app plans. Via the API (for developers), the small model is priced at $1 per million tokens in and $5 per million tokens out. That compares to Sonnet 4.5 with $3 per million inputs and $15 per million output tokens, and Opus 4.1 with $15 per million inputs and $75 per million output tokens.

The model serves as a cheaper replacement for two older models, Haiku 3.5 and Sonnet 4. “Users who rely on AI for real-time, low-latency tasks, such as chat assistants, customer service agents, or pair programming, will appreciate Haiku 4.5’s combination of high intelligence and remarkable speed,” Anthropic writes.

Claude 4.5 Haiku answers the classic Ars Technica AI question: "Would the color be called 'magenta' if the city of Magenta did not exist?"

Claude 4.5 Haiku answers the classic Ars Technica AI question: “Would the color be called ‘magenta’ if the city of Magenta didn’t exist?”

In SWE-bench VerifiedA test that measures performance on coding tasks, Haiku 4.5 scored 73.3 percent compared to Sonnet 4’s similar performance level (72.7 percent). The model also reportedly outperforms the Sonnet 4 in certain tasks such as using computers, according to Anthropic benchmarks. Claude Sonnet 4.5, released in late September, remains Anthropic’s cutting-edge model and what the company calls “the best encoding model available.”

Haiku 4.5 also surprisingly comes close to what OpenAI’s GPT-5 can achieve on this particular set of benchmarks (as seen in the graph above), although since the results are self-reported and potentially cherry-picked to match a model’s strengths, they should always be taken with a grain of salt.

Still, creating a small, capable coding model can have unexpected benefits for agent coding setups like Claude Code. Anthropic designed Haiku 4.5 to work alongside Sonnet 4.5 in multi-model workflows. In such a setup, Anthropic says, Sonnet 4.5 could break down complex problems into multi-step plans and then coordinate multiple Haiku 4.5 instances to complete subtasks in parallel, such as separating workers to get things done faster.

For more details on the new model, Anthropic released a system card and documentation for developers.

#Anthropics #Claude #Haiku #matches #Mays #frontier #model #fraction #cost

Leave a Reply

Your email address will not be published. Required fields are marked *