GPT-5.2 Codex: benchmark scores, price and ranking
GPT-5.2 Codex from OpenAI ranks #76 of 743 LLMs on the LLMs Tiger aggregate benchmark score (52.2), based on 12 benchmark results captured up to 2026-10-13.
- Organisation
- OpenAI
- Type
- Frontier model
- Open weights
- No
- Released
- 2025-12-18 (source: epoch)
- Context window
- 272,000 tokens
- Cheapest API price
- $1.75 in / $14.00 out per million tokens (via azure, observed 2026-10-13T19:47Z)
- Aggregate rank
- #76 of 743 (score 52.2)
Relative to other models, GPT-5.2 Codex is strongest on LiveBench Coding (#1 of 17) and weakest on SWE-bench Multilingual (#9 of 11).
Across the 12 benchmarks where at least ten models are scored, GPT-5.2 Codex is in the top 10% on 2 and in the top half on 8.
Its cheapest listed input price, $1.75 per million tokens, is lower than that of 24% of the 427 models with a listed price.
GPT-5.2 Codex benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| LiveBench Coding | 83.6% | #1 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| Terminal-Bench 2.0 | 66.5% | #2 of 28 | Epoch AI Benchmarking Hub | 2026-10-13 |
| ALE-Bench | 1,300 | #13 of 79 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SWE-bench Verified | 72.8% | #15 of 73 | SWE-bench Verified leaderboard | 2026-10-13 |
| LiveBench Agentic Coding | 49.4% | #5 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Data Analysis | 78.2% | #5 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench IF | 66.4% | #8 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Mathematics | 88.8% | #8 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Reasoning | 77.7% | #12 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| WebDev Arena | 1,339 | #48 of 67 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LiveBench Language | 73.7% | #13 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| SWE-bench Multilingual | 66.3% | #9 of 11 | SWE-bench Multilingual leaderboard | 2026-10-13 |
About GPT-5.2 Codex
GPT-5.2 Codex is a frontier model from OpenAI. It was released on 2025-12-18, according to Epoch AI. It accepts up to 272,000 tokens of input at once. The cheapest API listing we track charges $1.75 per million input tokens and $14.00 per million output tokens, through azure. We hold 12 benchmark results for it across 3 areas, and 2 are in the top tenth of their boards.
Where GPT-5.2 Codex stands
On our aggregate score, GPT-5.2 Codex is #76 of 743 ranked models, in the top quarter. Its score is 52.2; the scale is centred on 50, so it sits 2.2 points above the typical model. That beats 90% of the ranked field. Among the 88 ranked OpenAI models it is #12. The next one up is GPT-5.1 (high) at 52.7. The next one down is gpt-5.2-2025-12-11-high at 52.1.
GPT-5.2 Codex, benchmark by benchmark
LiveBench Coding is where GPT-5.2 Codex does best: 83.6%, first of 17. The next result is GPT-5.5 (xhigh) at 82.1%, 1.5 points behind. It covers LiveBench release 2026_06_25. The board median is 76.5%, so GPT-5.2 Codex is 7.2 points above it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
On Terminal-Bench 2.0, GPT-5.2 Codex scores 66.5% and ranks #2 of 28. Only one model scores higher, led by Gemini 3 Pro Preview at 69.4%, which is 2.9 points ahead. It covers real terminal tasks (agent + model). The board median is 37.3%, so GPT-5.2 Codex is 29.2 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-5.2 Codex manages 1,300 on ALE-Bench, good for #13 of 79. GPT-6 Astra (max) leads at 2,951, 1,651 points further up, and 84% of models sit below. The results either side are DeepSeek V4 Flash 0731 (max) (1,306) and Gemini 3.8 Flash (high) (1,270). It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so GPT-5.2 Codex is 536 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
SWE-bench Verified puts GPT-5.2 Codex at 72.8%, #15 of 73. That beats 79% of the models tested, but 10.7 points short of Claude Opus 4.7 (max) (83.5%). The results either side are GLM-5 (72.8%) and Claude 4 Sonnet (70.8%). It covers real GitHub issues resolved. The board median is 56.4%, so GPT-5.2 Codex is 16.4 points above it. Source: SWE-bench Verified leaderboard, captured 2026-10-13.
On LiveBench Agentic Coding, GPT-5.2 Codex reaches 49.4%. It ranks #5 of 17, ahead of 71% of the field, and 6.7 points behind claude-opus-4-8-xhigh-effort at 56.1%. The results either side are gpt-5.2-2025-12-11-high (50.3%) and gemini-3.5-flash-high (49.0%). It covers LiveBench release 2026_06_25. The board median is 45.8%, so GPT-5.2 Codex is 3.6 points above it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
GPT-5.2 Codex manages 78.2% on LiveBench Data Analysis, good for #5 of 17. GPT-5.5 (xhigh) leads at 81.6%, 3.4 points further up, and 71% of models sit below. The results either side are claude-opus-4-8-xhigh-effort (78.3%) and gpt-5.2-2025-12-11-high (78.2%). It covers LiveBench release 2026_06_25. The board median is 71.8%, so GPT-5.2 Codex is 6.4 points above it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
GPT-5.2 Codex lands below the middle on LiveBench Reasoning: 77.7%, #12 of 17, with 65% of the field scoring higher. claude-opus-4-8-xhigh-effort leads at 89.7%; the closest result above is kimi-k2.6-thinking at 79.4%. It covers LiveBench release 2026_06_25. The board median is 81.1%, so GPT-5.2 Codex is 3.4 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
GPT-5.2 Codex lands below the middle on WebDev Arena: 1,339, #48 of 67, with 70% of the field scoring higher. Claude Opus 5.5 (max) leads at 1,820; the closest result above is GLM-4.6 at 1,341. It covers building web apps, crowd-voted (Elo). The board median is 1,433, so GPT-5.2 Codex is 95 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-5.2 Codex lands below the middle on LiveBench Language: 73.7%, #13 of 17, with 71% of the field scoring higher. GPT-5.5 (xhigh) leads at 87.4%; the closest result above is qwen3.6-plus at 75.0%. It covers LiveBench release 2026_06_25. The board median is 77.9%, so GPT-5.2 Codex is 4.2 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
SWE-bench Multilingual is a weak spot. GPT-5.2 Codex scores 66.3%, #9 of 11, behind 73% of the models measured. The leader, Gemini 3 Flash, reaches 72.7%. It covers real GitHub issues resolved. The board median is 68.3%, so GPT-5.2 Codex is 2.0 points below it. Source: SWE-bench Multilingual leaderboard, captured 2026-10-13.
The full table above also lists the other 2 results, including boards with fewer than ten models, where a rank says little.
Strengths and weaknesses
Software engineering: solid, ahead of 75% of the field on average across 6 benchmarks. Its best board here is LiveBench Coding (#1 of 17) and its weakest is WebDev Arena (#48 of 67). Reasoning, maths and science: uneven, ahead of 41% of the field on average across 2 benchmarks. Its best board here is LiveBench Mathematics (#8 of 17) and its weakest is LiveBench Reasoning (#12 of 17). General knowledge and chat: uneven, ahead of 41% of the field on average across 4 benchmarks. Its best board here is LiveBench Data Analysis (#5 of 17) and its weakest is SWE-bench Multilingual (#9 of 11).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so GPT-5.2 Codex costs more than the typical listing. Models that score within 1.5 points or better and cost less: GLM-5.2 (high) ($0.060, 51.6), DeepSeek V4 Flash 0731 (max) ($0.070, 53.1), Qwen3-Coder-30B-A3B-Instruct ($0.070, 52.7). Similar scores cost more on Grok 4.6 (xhigh) ($2.00) and Grok 4.7 (xhigh) ($2.00).
Release history
The OpenAI model released before it was GPT-5.2 (xhigh) on 2025-12-11, which scores 51.9. The next one was GPT-5.3 Codex (xhigh) on 2026-02-05, at 54.3.
Reading these numbers
The 12 scores on this page come from 4 sources: LiveBench 2026_06_25 official CSV (7), Epoch AI Benchmarking Hub (3), SWE-bench Multilingual leaderboard (1), SWE-bench Verified leaderboard (1). 7 of them come from an official leaderboard and 5 from datasets or third-party runs.
Models ranked nearby
- #73 Darwin-397B-ZTC (52.2)
- #74 gemini-3.5-flash-high (52.2)
- #75 deepseek-v4.1-flash-max (52.2)
- #77 Ornith-1.5-397B (52.2)
- #78 Claude Code + Opus 5 (52.1)
- #79 Qwen3.8-2.4T-A95B (52.1)
More from OpenAI
- GPT-6 Astra (max) (#2, 64.0)
- GPT-6.1 Sol (max) (#5, 60.6)
- GPT-5.6 Sol (max) (#6, 60.1)
- GPT-6 Sol (max) (#7, 60.0)
- GPT-5.5 (xhigh) (#11, 58.0)
- GPT-5.6 Terra (max) (#12, 57.8)
Compare it on the live leaderboard → · How the aggregate score works