History
Haiku 4.5 makes the smaller-model choice concrete
Anthropic introduced a small model with performance compared to Sonnet 4 level.

On 15 October 2025, Anthropic introduced Claude Haiku 4.5. The announcement describes it as a small model at Sonnet 4 level. The announcement therefore concerned the relationship between a model’s class and what it could do. It was an update within the Haiku series, aimed at a different place in the range from the larger models. Source.

Its significance lay in choosing by task. The announcement was a reason to reconsider which model was needed for which work. A general comparison between models does not establish how much a particular business will benefit. That requires making the business’s own task visible.

For a small or medium-sized business, a manageable trial is sorting public texts into categories you choose yourself. Define the correct classification beforehand and include an ambiguous case. Have a smaller and a larger model process the same collection. Compare errors, the review needed and the trial’s total costs. Then choose based on the result for that task. Keep the original examples so you can later see whether another update changes the outcome.