Per-token prices fell, but agents spend far more tokens per task. Budget by task, not by token.
Claude Sonnet 5.5 (September 28): $2 per million input tokens and $10 per million output. Anthropic says it runs 30%+ faster than Sonnet 5 and costs up to 30% less for most work. It's generally available in GitHub Copilot.
GPT-6.1 Sol (OpenAI DevDay, September 29): the same $2 and $10, pitched as "near-Astra intelligence for a fifth of the price." It's in the API as gpt-6.1-sol, in Codex, and in Copilot for Pro+, Max, Business and Enterprise.
