Model2h ago
Anthropic Debuts Claude Opus 5
A 30.2 percent score on the ARC-AGI-3 benchmark puts Claude Opus 5 nearly four times ahead of GPT-5.6 Sol. Anthropic claims the model matches Fable 5 performance while cutting token costs by half. This shift lowers the price of high-end reasoning. Developers can now deploy flagship capabilities with significantly reduced overhead.