Anthropic on Wednesday launched Claude Haiku 5.5, the first new version of its smallest and most affordable model in nearly a year.

That’s Anthropic’s third 5.5 model in a month, though what’s still missing is Fable 5.5. That model, however, will likely go through a much longer review process, so it’s no surprise its launch is taking a bit longer.

Like with previous iterations, the company describes Haiku 5.5 as its “fastest and most efficient model,” but this time, it is also much cheaper. While Haiku 4.5 cost $1/$5 per million input/output tokens, Anthropic is bringing the price down to $0.10/$0.50 for requests under 100,000 tokens.

Anthropic describes Haiku 5.5 as its “fastest and most efficient model,” but this time, it is also much cheaper.

For larger requests, the new price is $0.50/$2.50 (Haiku 4.5 costs the same, no matter the number of tokens in a request), but Anthropic notes that about 90% of requests to Haiku 4.5 fell into the lower-priced category (which is likely a function of the kind of work that developers have sent to Haiku in the past).

Those are 90% and 50% cuts to the per-token prices, respectively. Anthropic puts the average savings at around 75%. That accounts for a mix of requests and an updated tokenizer that uses slightly more tokens per task.

This new version is the first Haiku model with effort controls (the default is medium), giving developers control over how many tokens the model uses on a given task.

The use case for these small models has always been to handle high-volume tasks like summarization, classification, and routing.

Source link