Anthropic's latest AI model is here. But it's more expensive than the previous generation, and unlike Anthropic's other models, it still can't analyze images.
The Claude 3.5 Haiku, announced by Anthropic last month, matches or outperforms Anthropic's flagship Claude 3 Opus in certain benchmarks. Claude 3.5 Haiku, available through Anthropic's API and numerous third-party platforms including AWS Bedrock, can help with coding suggestions, data extraction and labeling, and content moderation, Anthropic said.
Anthropic previously hinted that the Claude 3.5 Haiku will be priced the same as its predecessor, the Claude 3 Haiku. Now that tone is changing.
“During final testing, the Haiku outperformed its previous flagship model, the Claude 3 Opus, on many benchmarks at a fraction of the cost.As a result, we reduced the price of the Claude 3.5 Haiku to reflect its increased intelligence. We’ve raised it,” Anthropic wrote in a post about X.
Pricing for Claude 3.5 Haiku starts at $1 per million input tokens (approximately 750,000 words) and $5 per million output tokens. In comparison, Claude 3 Haiku's 25 cents per million input tokens and $1.25 per million output tokens is a 4x increase.
Claude 3.5 Haiku also lacks image analysis capabilities, making it less capable in at least one respect than Claude 3 Haiku. Alex Albert, head of developer relations at Anthropic, told X that Claude 3 Haiku will continue to be available to “users who require maximum cost efficiency and image processing.”
Indeed, Claude 3.5 Haiku can output longer chunks of text than Claude 3 Haiku, and the model has a more recent knowledge cutoff (i.e., can see more recent events). Still, it's rare for AI vendors to increase the cost of models in the series, raising questions about Anthropic's future pricing strategy.