GPT-5.2
Multimodal
#1GPQA Diamond
#1GDPVal
#2MCP-Atlas
+4 more
by OpenAI
+
+
+
+
About
GPT-5.2, released by OpenAI on December 11, 2025, is a large language model from the GPT-5 family that improves on GPT-5 in general intelligence, long-context understanding, agentic tool-calling, and vision. It features a 400K token context window, 128K maximum output tokens, and a knowledge cutoff of August 2025. GPT-5.2 targets long-context coding tasks, extended document analysis, and complex agentic workflows requiring reliable instruction following.
+
+
+
+
Timeline
ReleasedDec 11, 2025
Knowledge CutoffAug 1, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown
Overall Performance
9 benchmarks
Average Score
56.9%
Best Score
92.4%
High Performers (80%+)
2Top Categories
Science
92.4%
Coding
61.7%
Tool Use
60.6%
Finance
58.5%
Agents
49.7%
+
+
+
+
All Benchmark Results for GPT-5.2
Complete list of benchmark scores with detailed information
| GPQA Diamond | Science | 92.40 | 92.4% | Unverified | |
| SWE Bench Verified | Coding | 80.00 | 80.0% | Unverified | |
| MCP-Atlas | Tool Use | 60.57 | 60.6% | Unverified | |
| Finance Agent | Finance | 58.53 | 58.5% | Unverified | |
| Terminal Bench 2.0 | Coding | 54.00 | 54.0% | Unverified | |
| SWE-rebench | Coding | 51.00 | 51.0% | Unverified | |
| GDPVal | Agents | 49.70 | 49.7% | Unverified | |
| ARC-AGI-2 | Reasoning | 38.50 | 38.5% | Unverified | |
| Humanity's Last Exam | Reasoning | 27.80 | 27.8% | Unverified |