Anthropic Launches Claude 3 Model Family With Three Tiers of AI
Anthropic introduces the Claude 3 model family, featuring Haiku, Sonnet, and Opus to provide users with flexible options for intelligence, speed, and cost. The new models set industry benchmarks in reasoning, mathematics, and vision capabilities.
Anthropic announces the Claude 3 model family, setting new industry benchmarks across a wide range of cognitive tasks. This release includes three state-of-the-art models in ascending order of capability: Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus. Users can now select the optimal balance of intelligence, speed, and cost for their specific applications, with Opus and Sonnet currently available in claude.ai and the generally available API across 159 countries.
Opus, the most intelligent model in the family, outperforms peer models on common evaluation benchmarks including undergraduate-level expert knowledge and graduate-level reasoning. It exhibits near-human levels of comprehension and fluency on complex tasks, while all Claude 3 models show increased capabilities in analysis, forecasting, code generation, and non-English languages. The models also feature near-instant results, with Haiku reading a dense research paper with charts in under three seconds and Sonnet operating twice as fast as Claude 2.1.
The Claude 3 models possess sophisticated vision capabilities on par with other leading AI systems. They process a wide variety of visual formats, including photos, charts, graphs, and technical diagrams. Anthropic expresses particular excitement about bringing this new visual modality to enterprise customers who have eagerly awaited advanced document and image processing features.