whatmodel.sh / GPT-5.2-Codex

GPT-5.2-Codex

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

GPT-5.2-Codex is a closed-weights language model from OpenAI, released January 2026.

Catalogue identity notes

Comparison-only. Arena entry 'gpt-5.2-codex' (webdev). Released 2026-01: OpenAI API changelog, Jan 14 (2026): 'Released gpt-5.2-codex to the Responses API'. Shut down 2026-07-23.

68.5±1.7comparison only: no current run option verifiedfrom 5 benchmark results across 3 sources

$ how it compares

The LLMs ranked either side of GPT-5.2-Codex, out of 828 with a score, then more from OpenAI.

#modelscorewhereRAMreleased
111Qwen3.5 27B Alibaba69±0api $0.84/M +local~18GB2026-02
112GLM-5V-Turbo Z.ai69±2api $1.90/M2026-04
113GPT-5.2-Codex OpenAI69±2comparison only2026-01
114LongCat-Flash-Thinking (560B-A27B) Meituan69±1comparison only2025-09
115Step 3.5 Flash (196B-A11B) StepFun68±1api $0.15/M +local~118GB2026-02
more from OpenAI
1GPT-6 Astra OpenAI87±1api $20.00/M2026-09
3GPT-6.1 Sol OpenAI85±1api $4.00/M2026-09
7GPT-6 Sol OpenAI83±1api $4.00/M2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▂▆█▆▅▄▄▃▂▁ #63 of 340 · 52% of its score · 4 tests

Beats Gemini 3.6 Flash, GPT-5 Pro, LongCat 2.0 (1.6T-A48B) here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Terminal-Bench 2.0 lab64.0% ✎██···██··███ beats 3/6very hard · avg 14%▇▇▇▇▇▇▇▇ 21%at its level
OSWorld-Verified lab38.2% ✎█▄·····▄·▄▄▄ beats 0/6hard · avg 17%▇▇▇▇▇ 12%at its level
Vibe Code Bench (test split, 50 web apps)37.9% ✎▃··▃·▃██▃▃▃█ beats 10/14very hard · avg 2.3%▇▇▇▇▇ 12%sharp · noisy
SWE-bench Pro lab56.4% ✎▄·····▄·▄▄▄█ beats 3/6hard · avg 28%▇▇▇ 7.4%blunt · at its level
coding▁▁▁▃▅▇█▇▅▄▃▁ #128 of 524 · 66% of its score · 3 tests

Beats Qwen3.5 Plus here, though they rank higher overall. Trails Llama 3.1 Nemotron Ultra 253B v1, Solar Pro 4, DeepSeek R1 0528 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena WebDev (Code Arena)1338.9 ✎▁▂▄▃▅█▄▇▆▄▁▁ beats 25/119avg 981▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 47%sharp
Vibe Code Bench (test split, 50 web apps)37.9% ✎▃··▃·▃██▃▃▃█ beats 10/14very hard · avg 2.3%▇▇▇▇▇ 12%sharp · noisy
SWE-bench Pro lab56.4% ✎▄·····▄·▄▄▄█ beats 3/6hard · avg 28%▇▇▇ 7.4%blunt · at its level

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · LMArena leaderboard dataset 47%, OpenAI vendor-reported benchm… 40%, Vibe Code Bench (Vals AI, arX… 12%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset1▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 47%CC BY 4.0
OpenAI vendor-reported benchmark scores for its own models (openai.com announcement posts and Deployment Safety Hub system cards) · lab-reported3▇▇▇▇▇▇▇▇▇▇▇▇ 40%Factual vendor-reported score (not licensed content)
Vibe Code Bench (Vals AI, arXiv:2603.04601v3)1▇▇▇ 12%CC-BY-4.0
prior · pull toward the middle▇ 1%
weighted average= 68.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json