Atlas

Models

← All models

GPT-5.2

OpenAI · 2025-12-11

GPT-5.2 is OpenAI's reasoning model for complex professional work and the API counterpart to ChatGPT GPT-5.2 Thinking, with configurable reasoning effort. It improved on GPT-5.1 in reasoning and agentic work and was later followed by the separately released GPT-5.2-Codex variant optimized for agentic coding. Its only published snapshot, and the gpt-5.2 alias's default, is gpt-5.2-2025-12-11.

Benchmark scores

BenchmarkScore
A-Fantasia51.3
A-Fantasia (Valid Attempts with Retries)51.3
A-Fantasia Backwards Spelling37.0
A-Fantasia Chess42.0
A-Fantasia Chess (Valid Attempts with Retries)42.0
A-Fantasia Cube (Valid Attempts with Retries)75.0
A-Fantasia Cube Rotation75.0
A-Fantasia Spelling (Valid Attempts with Retries)37.0
AA-LCR72.7
AA-LCR v1.182.7
AA-Omniscience Index0.3
Agent Red Teaming k=100 (CAIS)68.8
AIME 2024 (avg@8)97.1
AIME 2024-202596.9
AIME 2025100.0
AIME 2025 (Artificial Analysis avg@10)99.0
AIME 2025 (avg@8)96.7
AIME 2025 (MathArena repeated-run mean)100.0
AIME 202698.3
ALE-Bench1715
AlgoTune2.1
Apex (MathArena)13.5
APEX (Mercor)11.9
Apex Shortlist79.3
APEX-Agents32.4
APEX-v1-extended64.7
ARC-AGI-1 Semi-Private78.7
ARC-AGI-243.3
Arena Code WebDev — Brand & Marketing1395
Arena Code WebDev — Consumer Product1418
Loading Atlas data…