OpenAI and Anthropic Cut LLM Prices by Up to 80% in Months — Why It Happened
minimaxir · x · 2026-10-26
- Max Woolf's blog post documents a sharp LLM price war: within a few months, OpenAI and Anthropic cut prices by up to 80%, against his expectations.
- Key pricing timeline (input/output/cache-read per MTok): GPT-5.6 Sol launched at $5/$30/$0.50; small model Luna at $1/$6/$0.10; Claude Sonnet 5 debuted at a 'temporary' $2/$10/$0.20; Claude Opus 5 at $5/$25/$0.50.
- On July 30 OpenAI slashed Luna's price 80% to $0.20/$1.20/$0.02 — its own chart shows Luna would otherwise have lost to DeepSeek V4 Pro on quality-vs-cost-per-task.
- Sonnet 5's launch price later became permanent, Sol got a 20% cut to $4/$20/$0.40, and Fable 5.1 arrived at $10/$50/$0.25 — competitive pressure, especially from cheaper rivals like DeepSeek, is driving the declines.
More from Models
- ChatGPT interrupts itself explaining tactile communication, users note rising bugginess — Angaisb_ · 2026-10-26
- Rumor: Gemini 4 reliably disagrees with every other LLM, unlike blending frontier models — vitaliychiley · 2026-10-26
- Anthropic quietly restricts Claude's reverse-engineering ability, sparking open-model debate — jdjohnson · 2026-10-26
- Hands-on: Opus 5.5 shows great "taste", Gemini 4 shines at search — apples_jimmy · 2026-10-26
- Codex Security includes $500 in non-expiring free credits for GitHub repo scans — ChrisUniverse · 2026-10-26
- NVIDIA highlights Baseten x CodeRabbit post-training open models to frontier level — baseten · 2026-10-26