Anthropic's fastest AI model now available to all Claude users
Anthropic has officially launched its fastest model, Claude 3.5 Haiku, for general users via the Claude chatbot on web and mobile apps. Previously exclusive to developers via Anthropic's API since October 2024, this compact yet speedy model has been recognized for outperforming larger models on key benchmarks while maintaining a competitive price point. The quiet release follows major updates from AI rivals OpenAI and Google, which recently launched o1 and Gemini 2.0 models, respectively.
Claude 3.5 haiku's features and limitations
Claude 3.5 Haiku is optimized for real-time tasks such as handling large datasets and analyzing financial documents, with a 200,000-token context window. This exceeds OpenAI's GPT-4o's 128,000-token window. The model improves Claude chatbot's functionality by enabling image and file attachment analysis. It also works with Claude Artifacts, an interactive sidebar for real-time content refinement launched in June 2024. However, it misses web browsing and image generation capabilities provided by competitors.
Access and subscription details for Claude 3.5 Haiku
While users can use Claude 3.5 Haiku for free, they will have a variable daily message limit depending on server demand. The free tier gives some 10 exchanges before hitting Anthropic's quota, which resets every day. A $20 per month subscription to Claude Pro plan provides more usage and other benefits like up to five times the free tier's usage, priority access during high-traffic periods, early access to new features, and access to additional models like Claude 3 Opus.
Claude 3.5 haiku's performance and cost on API
On the API, Claude 3.5 Haiku is also cost-effective, starting at just $0.80 per million input tokens and $4 per million output tokens. Developers can further reduce costs using prompt caching and the Message Batches API. In benchmark testing, Haiku scored 40.6% on SWE-bench Verified, showcasing its strength in tasks requiring intelligence and speed. Despite limitations like daily message caps and lack of certain features, it remains a powerful tool for tasks needing speed and precision.
