Claude Opus 4.7 — same per-token price, but the new tokenizer raises real cost ~35%

Anthropic shipped Opus 4.7 at unchanged $5/$25 per 1M, paired with Claude Design (visual collaboration). Most coverage missed the tokenizer change — Opus 4.7 generates ~35% more tokens for the same prompts as 4.6. Effective cost rose without the price tag moving.

Verdict: Frontier reasoning still cloud-only; new tokenizer adds ~35% effective cost

The take

Opus 4.7 generally available April 16. Same headline pricing as 4.6 ($5 input / $25 output per 1M). Improvements claimed in software engineering and complex long-running coding tasks; vision sees images at higher resolution. Anthropic Labs also launched Claude Design, a Figma-adjacent visual collaboration product.

The story most coverage missed: Opus 4.7 ships a new tokenizer that produces ~35% more tokens than 4.6 for the same prompts and outputs. That means the unchanged per-token price translates to a ~35% effective cost increase in real-world use. Multiple cost-tracking sources flagged this within the first week.

Why it matters for local-vs-cloud math: if you've been running Opus 4.6 API at, say, $200/month, expect that bill to land at ~$270/month on 4.7 with the same workload. The cost calculator on this site uses the corrected effective rate (perMillion 15) — but anyone running the math against the headline 4.6 numbers from a few months ago will be surprised.

Local-side, nothing changes today: Opus 4.7 is hosted-only, has no open-weight equivalent, and the frontier reasoning gap to local opens (Qwen 3.6-35B-A3B, GLM-5.1, Kimi K2.6) remains real for hard problems. If you need Opus-class reasoning, plan on the cloud bill being a third higher than your last invoice suggested.

Sources

Next step

Try this in the planner→