Anthropic Claude Sonnet 4.5 vs OpenAI GPT-5.2-Codex
Side by side from live data: prices, limits, who serves each model and how reliably, and how they do on our chess arena.
In short
- GPT-5.2-Codex is 1.2× cheaper at a typical mix of 3 input tokens per output token.
- Claude Sonnet 4.5 accepts longer prompts: 1M vs 400K tokens.
- GPT-5.2-Codex is the newer model (released 2026-01-14).
| Claude Sonnet 4.5 | GPT-5.2-Codex | |
|---|---|---|
| Vendor | Anthropic | OpenAI |
| Released | 2025-09-29 | 2026-01-14 |
| Input / 1M | $3.00 | $1.75 |
| Output / 1M | $15.00 | $14.00 |
| Cached input / 1M | $0.30 | $0.175 |
| Blended / 1M (3:1) | $6.00 | $4.81 |
| Context | 1M | 400K |
| Max output | 64K | 128K |
| Accepts | text, image, file | text, image |
| Reasoning | Yes | Yes |
| Providers on OpenRouter | 7 | 1 |
| Cheapest provider, output | $15.00 | $14.00 |
| Best provider uptime, 24 h | 100% | 100% |
| Chess arena rating | not playing | not playing |
| Vendor API status | Operational | Operational |
Cost calculator
Claude Sonnet 4.5
$405 per month$13.50 per day
GPT-5.2-Codex
$315 per month$10.50 per day
GPT-5.2-Codex saves $90.00 a month on this workload.
At list prices per 1M tokens; prompt caching and batch discounts are not included.
Compare Claude Sonnet 4.5 with