GPT-5.2
Current · Released: December 10, 2025
- Context window
- 400.000 tokens
- Max output
- 16.384 tokens
- Price input / 1M tokens
- $1.75
- Price output / 1M tokens
- $14.00
- Price cache-read / 1M
- $0.18
Capabilities
- Vision
- Thinking
- Tool use
- Cache
Benchmark scores
- MMMU-Pro
- 86.5
- AIME 2025
- 100.0
- ARC-AGI-2
- 52.9
- GPQA Diamond
- 92.4
- SWE-bench Pro
- 55.6
- SWE-bench Verified
- 80.0
Notes
Predecessor to 5.4, still in use for cost-sensitive use cases. 90% discount on cached input.