- Published on
OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?
- Authors
- Name
- DeepHorus
OpenAI and Anthropic have cut flagship LLM prices by up to 80% in months, reshaping agentic-coding economics. On June 26, 2026, OpenAI launched GPT-5.6 Sol at $5/$30/$0.50 per million tokens and Luna at $1/$6/$0.10. Anthropic followed on June 30 with Claude Sonnet 5 at $2/$10/$0.20, initially labeled temporary launch pricing.
Discounts soon became permanent and broader. On July 24, Anthropic released Claude Opus 5 at $5/$25/$0.50. Six days later, OpenAI cut GPT-5.6 Luna by 80% to $0.20/$1.20/$0.02, and Anthropic made Sonnet 5’s low price permanent on August 10.
On August 21, OpenAI discounted GPT-5.6 Sol by about 20% to $4/$20/$0.40 for three months. Cache-read pricing then became the decisive battlefield. On September 1, Anthropic released Fable 5.1 at $10/$50/$0.25, cutting cache-read costs by 75%.
Cache reads account for 96%-99% of token spending in agentic coding and hit 90%-100% of the time, so that cut translates to roughly 70%-75% lower total usage cost. OpenAI responded with GPT-6 Astra on September 3 at $10/$50/$1, matching Fable 5.1’s performance but not its cache discount. On September 22, OpenAI released GPT-6 Sol at $2/$10/$0.20 and GPT-6 Luna at $0.10/$0.50/$0.01.
That same day, Anthropic launched Claude Opus 5.5 at $4/$20/$0.20, citing lower serving compute and delivering a 40%-60% total cost reduction. Performance feedback was mixed: Sonnet 5’s reception was divided, Opus 5 drew mixed real-world reviews, and GPT-6 Sol was a dud. A week later, OpenAI announced GPT-6.1 Sol at $2/$10/$0.10, halving cache-read costs to match Anthropic’s 5%-of-input cache pricing.
GPT-6.1 Sol’s performance was comparable to Astra, and the Sol family became cumulatively 75% cheaper. The cuts also apply to subscription agentic harnesses such as Codex and Claude Code. Competition now centers on total agent cost and cache architecture, not headline token prices; further cuts may follow if serving efficiency improves, though sustainability is uncertain.
