Claude Sonnet 4.6 vs GPT-5.4
A direct API-pricing comparison between Claude Sonnet 4.6 and GPT-5.4. Below: every cost dimension side by side, then the bottom line for a typical workload.
Bottom line
For a typical 5K-in / 1K-out request, GPT-5.4 is the cheaper choice at $0.028 per call — 1.1× less than Claude Sonnet 4.6 at $0.030.
| Claude Sonnet 4.6Anthropic | GPT-5.4OpenAI | |
|---|---|---|
| Input / 1M tokens | $3.00 | $2.50 |
| Output / 1M tokens | $15.00 | $15.00 |
| Cached input / 1M | $0.300 | — |
| Context window | 200K | 400K |
| Provider | Anthropic | OpenAI |
Green = cheaper on that line. Prices are estimates from public pricing pages; verify against Anthropic and OpenAI before relying on them.
Which is cheaper, Claude Sonnet 4.6 or GPT-5.4?
On output tokens — usually the dominant cost — Claude Sonnet 4.6 charges $15.00 per million versus $15.00 for GPT-5.4. The right pick depends on your input/output ratio and how much you can serve from cache. Use the calculator to model your exact traffic.