Anthropic · 2025 · 11 · 24 · Model · ~2 min read
Anthropic launched Claude Opus 4.5
What's actually new
- Top of leaderboards at launch. Held the lead on most independent reasoning and coding tests for several weeks.
- Price cut. Lower input and output token cost than Opus 4 — Anthropic's pitch was 'flagship quality without flagship price'.
- Stronger on long agent runs. Multi-hour autonomous tasks held up better than Opus 4 on independent reliability tests.
If you want more
Worth knowing
- 'Best in the world' lasted ~2 months before GPT-5.2 and Gemini 3.1 Pro retook spots.
- Still not cheap. 'Cut' is relative — Opus 4.5 remains the most expensive top-tier mainstream model on a per-call basis.
Who should care
Teams running production AI agents who care about reliability over speed. Researchers needing the strongest reasoning available. Developers spending real money on Claude API and watching the cost line.
What to do about it
If you're already on Opus 4 in production, switch to 4.5 — same code, lower bill, better output. If you're on Sonnet 4.5 and the work doesn't fail, stay there — Opus 4.5 is overkill for most tasks.
Honest take
Opus 4.5 quietly reset the top of the price chart. Anthropic spent 2025 proving that the best model doesn't have to be the most expensive — and the price cut was their signal that the economics of frontier AI were starting to bend. Whether that holds through 2026 depends on how much each lab is willing to spend on training runs vs how much they need to charge to recoup it. The era of unprofitable frontier pricing isn't over.
Sources
Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com