Anthropic's Haiku 5.5: 75% Cheaper, 2.5x Faster

🔥 What happened Anthropic dropped Claude Haiku 5.5, its fastest and cheapest small model yet. It costs 75% less than Haiku 4.5 and crushes GPT-6 Luna on OSWorld 2.1 with 72.4% vs. 48.9%. Anthropic also halved Sonnet 5.5's cache-read pricing. 💡 Why it matters For agentic workloads, that means 2.5x faster inference per turn and 30% lower task-completion latency. At 8 million API calls per week — like AlphaSense runs — the savings compound fast. Small models just became economically viable for production subagents. ⚡ Our take Anthropic is playing the pricing game now, and playing it aggressively. If you're still running Haiku 4.5 in production, you're burning money. The real question: how long until OpenAI and Google respond?