GPT-5.2 (2025-12-11)
Dec 2025
GPT-5.2 (2025-12-11)Compare inon
Benchmarks of GPT-5.2 (2025-12-11)26
BenchmarkByScoreRank
Knowledge2
Reasoning1
Math4
Agents2
Legacy · saturated or no longer reported17
No benchmark matches
Rankingat xhigh effort?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGPT-6.1 Solmax+42%-17%$0.72Remove unsupported parameters+5
BetterClaude Haiku 5.5max+15%-96%$0.21Remove sampling parameters+5
OpenDeepSeek V4 Pro 0813max-4%-59%$0.67No image input+5CheaperClaude Sonnet 5.5max+23%-17%$5.46No forced tool use+5
CheaperGemini 4 Argonhigh+23%-17%*$1.62Synchronous output stops at 65,536+2ModelScoreCostPer taskWhat changes
GPT-6.1 Solmax+42%-17%$0.72Remove unsupported parameters+5Claude Sonnet 5.5max+23%-17%$5.46No forced tool use+5Gemini 4 Argonhigh+23%-17%*$1.62Synchronous output stops at 65,536+2Gemini 3.8 Flashhigh+21%-69%*$3.49Sampling parameters are ignored+1GPT-6 Solmax+17%-17%$1.04ModelScoreCostPer taskWhat changes
Claude Haiku 5.5max+15%-96%$0.21Remove sampling parameters+5GPT-6.1 Solmax+42%-17%$0.72Remove unsupported parameters+5GPT-6 Solmax+17%-17%$1.04Gemini 4 Argonhigh+23%-17%*$1.62Synchronous output stops at 65,536+2GPT-5.6 Solmax+23%+66%$1.99ModelScoreCostPer taskWhat changes
DeepSeek V4 Pro 0813maxOpen-4%-59%$0.67No image input+5DeepSeek V4.1 Flash 1017maxOpen-4%-89%$0.26Pass back reasoning_content · Chat Completions+4
Kimi K3highOpen-23%+25%$3 / $15Sampling parameters are fixed+3
GLM 5.3maxOpen-23%-55%$2.01No image input+4GLM 5.2maxOpen-54%-55%$1.28No image inputModelScoreCostPer taskWhat changes
GPT-6 Astramax+54%—$3.26Monitoring can return HTTP 403+2GPT-6.1 Solmax+42%—$0.72Remove unsupported parameters+5GPT-5.5xhigh+31%—$2.10Prompts with >272K input tokensGPT-5.5 Proxhigh+29%—$30 / $180GPT-5.6 Solmax+23%—$1.99GPT-5.4 Pro (2026-03-05)+23%—$30 / $180No strict JSONGPT-6 Solmax+17%—$1.04GPT-5.4 (2026-03-05)xhighsame—$2.50 / $15All OpenAI models →History
17 versionsBenchmarks
all categoriesChat & Knowledge
SimpleQA Verified v137.1#45
FrontierMath40.7#5
FrontierMath Tier 418.8#5
FrontierMath Tiers 1–3 (v2)67.4#28
FrontierMath Tier 4 (v2)31.7#29
Chess Puzzles48#15
GPQA Diamond · Epoch91.4#23
OTIS Mock AIME 2024 202596.1#31
Coding
SWE-rebench · all · December 1, 2025 to February 1, 202656.6#1
SWE-rebench · Go · December 1, 2025 to February 1, 20260#1
SWE-rebench · Java · December 1, 2025 to February 1, 20260#1
SWE-rebench · Rust · December 1, 2025 to February 1, 20260#1
SWE-rebench · Python · December 1, 2025 to February 1, 202656.6#1
SWE-rebench · TypeScript · December 1, 2025 to February 1, 20260#1
Agents & Tool Use
EBR Bench23#16
Furniture Assembly38.3#18
FAQ
How much does GPT-5.2 (2025-12-11) cost?›
$1.75 per 1M input tokens and $14 per 1M output tokens.
Is GPT-5.2 (2025-12-11) free to use?›
There is no free API; the cheapest paid price is $1.75 per 1M input tokens.
How good is GPT-5.2 (2025-12-11), and what is it best at?›
#45 of 84 in Chat & Knowledge, #14 of 24 in Coding, #7 of 39 in Agents & Tool Use.
Is GPT-5.2 (2025-12-11) better than GPT-5 Pro (2025-10-06)?›
On Chess Puzzles, GPT-5.2 (2025-12-11) scores 48 vs 48 — about the same.
How can I use GPT-5.2 (2025-12-11)?›
Through the OpenAI API.
What is the context window of GPT-5.2 (2025-12-11)?›
400K tokens.
When was GPT-5.2 (2025-12-11) released? Is there a newer version?›
Released 2025-12-11. It is the current version in its line.
Can GPT-5.2 (2025-12-11) read images?›
It accepts text and images as input.
What is GPT-5.2 (2025-12-11)?›
GPT-5.2 (2025-12-11) is a language model from OpenAI, released Dec 2025.
What are the best alternatives to GPT-5.2 (2025-12-11)?›
GPT-6.1 Sol (cheaper), Claude Haiku 5.5 (better), DeepSeek V4 Pro 0813 (open weights).