Claude 3 (Opus, Sonnet, Haiku): Anthropic challenges GPT-4
In one sentence Anthropic ships the Claude 3 family in three sizes, all with vision and a 200K-token context. According to Anthropic, Opus outperforms its peers on most common benchmarks, GPT-4 included. Opus and Sonnet available immediately, Haiku coming soon.
Anthropic ships Claude 3 in three versions: Opus (the most capable and expensive), Sonnet (balanced), Haiku (small, fast, cheap). All three read images, charts and documents.
The headline: according to Anthropic's tables, Opus beats GPT-4 on most widely used benchmarks. After a year in which GPT-4 was the undisputed reference, another model matches or surpasses it.
For developers, it's the moment "default = GPT-4" stops being an obvious choice.
Companies
Anthropic
Tools
Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku
Tags
Sources