Request a Pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT-5.2 Codex

OpenAI

Strongest in Instruction following (#38 of 62), weakest in Core abilities (#128 of 199). Above par in 0 of 7 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and GPT-6 Astra and ahead of NVIDIA Nemotron 3 Ultra 550B A55B and GLM-5.3 Flash.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Coding47.7#101/179
Agentic coding47.6#98/173
Reasoning49.6#104/195
Mathematics49.4#88/152
Core abilities46.9#128/199
Instruction following47.2#38/62
Language44.2#49/62

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for GPT-5.2 Codex, left for the other.

GPT-4o miniwon 5 · lost 0
Command Awon 5 · lost 0
Mistral Medium 3.5won 5 · lost 0
GPT-4.1 nanowon 5 · lost 0
Gemini 2.0 Flashwon 5 · lost 0
Inkling Smallwon 5 · lost 0
NVIDIA Nemotron 3.5 Lightning 30B A3Bwon 7 · lost 0
Qwen3.6 27Bwon 7 · lost 0
GPT-5.4 miniwon 7 · lost 0
Grok Build 0.1won 7 · lost 0
MiniMax M3won 6 · lost 1
Inklingwon 6 · lost 1
Gemini 3.5 Flash-Litewon 6 · lost 1
Grok 4.3won 6 · lost 1
GPT OSS 120Bwon 4 · lost 1
GPT-4.1 miniwon 4 · lost 1
Mistral Large 3won 4 · lost 1
Llama 4 Maverick Instructwon 4 · lost 1
Ox Alphawon 5 · lost 2
Mistral Large 4won 5 · lost 2
Kimi K2.6won 5 · lost 2
Step 5won 3 · lost 2
DeepSeek V3won 3 · lost 2
GPT-5.4 nanowon 4 · lost 3
Qwen3.6 Pluswon 4 · lost 3
Kimi K2.7 Codewon 4 · lost 3
GPT-5.6 Lunawon 4 · lost 3
GLM-5.3 Flashwon 4 · lost 3
NVIDIA Nemotron 3 Ultra 550B A55Bwon 4 · lost 3
Claude Opus 4.5won 3 · lost 4
Qwen3.7 Maxwon 3 · lost 4
Beam 1.0won 2 · lost 3
GLM-4.6won 2 · lost 3
Kimi K2won 2 · lost 3
GPT-4.1won 2 · lost 3
MiniMax M2.5won 2 · lost 3
Gemini 2.5 Flashwon 2 · lost 3
Claude Sonnet 4.5won 2 · lost 3
Grok 3won 2 · lost 3
Kimi K2 Thinkingwon 2 · lost 3
Grok 4 Fastwon 2 · lost 3
Gemini 3.1 Flash-Litewon 2 · lost 3
GPT-5 miniwon 2 · lost 3
Qwen3 Maxwon 2 · lost 3
Claude Opus 4won 2 · lost 3
Grok 4.20won 2 · lost 3
GLM-5won 2 · lost 3
Kimi K2.5won 2 · lost 3
Gemini 3 Prowon 2 · lost 3
Claude Haiku 5.5won 2 · lost 5
GPT-6 Lunawon 2 · lost 5
Claude Sonnet 5won 2 · lost 5
Claude Sonnet 4.6won 2 · lost 5
Qwen3.8 27Bwon 2 · lost 5
DeepSeek V4 Flashwon 2 · lost 5
DeepSeek V4 Flash Visionwon 2 · lost 5
Qwen3.8 Flash Nextwon 2 · lost 5
Gemini 3.1 Prowon 2 · lost 5
MiMo V2.6 Prowon 1 · lost 4
Qwen3 235B A22B Instructwon 1 · lost 4
O4 Miniwon 1 · lost 4
DeepSeek R1won 1 · lost 4
Gemini 2.5 Prowon 1 · lost 4
Claude Sonnet 4won 1 · lost 4
GLM-5.2won 1 · lost 6
Muse Spark 1.1won 1 · lost 6
GPT-5.2won 1 · lost 6
GPT-5.6 Terrawon 1 · lost 6
Claude Opus 5won 1 · lost 6
GLM-5.3won 1 · lost 6
Muse Spark 1.2won 1 · lost 6
Claude Opus 4.6won 1 · lost 6
Gemini 3.6 Flashwon 1 · lost 6
Gemini 3.5 Flashwon 1 · lost 6
Grok 4.5won 1 · lost 6
Claude Opus 5.5won 1 · lost 6
DeepSeek V3.2won 0 · lost 5
MiMo V2.6 Flashwon 0 · lost 5
Grok 3 Miniwon 0 · lost 5
O3won 0 · lost 5
GPT-5won 0 · lost 5
Kimi K3.1 Prowon 0 · lost 5
Grok 4won 0 · lost 5
Muse Wonder Fullwon 0 · lost 5
Gemini 4 Argonwon 0 · lost 5
Claude Fable 5.2won 0 · lost 5
DeepSeek V4 Prowon 0 · lost 7
Gemini 3.8 Flashwon 0 · lost 7
Grok 4.6won 0 · lost 7
GPT-6 Solwon 0 · lost 7
Qwen3.8 Maxwon 0 · lost 7
Claude Sonnet 5.5won 0 · lost 7
Claude Opus 4.7won 0 · lost 7
DeepSeek V4.1 Flashwon 0 · lost 7
GPT-5.4won 0 · lost 7
Kimi K3won 0 · lost 7
Grok 4.7won 0 · lost 7
GPT-5.6 Solwon 0 · lost 7
Muse Spark 1.3won 0 · lost 7
Muse Wonderwon 0 · lost 7
Claude Opus 4.8won 0 · lost 7
Claude Fable 5.1won 0 · lost 7
Gemini 3.7 Flashwon 0 · lost 7
GPT-5.5won 0 · lost 7
GPT-6.1 Solwon 0 · lost 7
GPT-6 Astrawon 0 · lost 7
Claude Fable 5won 0 · lost 7

§ 3 · Sources

Where the numbers come from

2 publications, 10 figures. Every one links to the page it was read from.

LiveBench 1.0 ↗8 measuresread 2026-11-10
LiveBench 1.0 74LiveBench 1.0 · Reasoning 77.7LiveBench 1.0 · Coding 83.6LiveBench 1.0 · Agentic Coding 49.4LiveBench 1.0 · Mathematics 88.8LiveBench 1.0 · Data Analysis 78.2LiveBench 1.0 · Language 73.7LiveBench 1.0 · Instruction Following 66.4
Vals.ai ↗2 measuresread 2026-11-10
Vals · SWE-bench Verified 72.4% (Mini-SWE-agent)Vals · Vibe Code Bench 37.91% (OpenHands)

Badge

PublicAI Index badge for GPT-5.2 Codex[![PublicAI Index](https://publicai.io/model-index/badge?model=gpt-5-2-codex)](https://publicai.io/model-index/m/gpt-5-2-codex)