Rankings / OpenAI
GPT 5.2 Codex
released 2025-12-11
BenchAtlas Index
as of 2026-07-14
xhigh effort
rank #49
8 families · 4 categories · medium
base configuration
rank #123
7 families · 4 categories · medium
Benchmark evidence
14 results
agentic coding
| Release 2026-06-25 tasks_counted=3 · livebench_version=2026-06-25 | 49.4 % | independent | LiveBench | |
| τ²-Bench | xhigh effort | 92.1 % | independent | Artificial Analysis |
coding
| Release 2026-06-25 tasks_counted=2 · livebench_version=2026-06-25 | 83.6 % | independent | LiveBench | |
| SciCode | xhigh effort | 54.6 % | independent | Artificial Analysis |
data analysis
| Release 2026-06-25 tasks_counted=3 · livebench_version=2026-06-25 | 78.2 % | independent | LiveBench |
external indices
| Intelligence Index v4.1 | xhigh effort | 40.1 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | xhigh effort | 89.9 % | independent | Artificial Analysis |
| Humanity's Last Exam implementation=artificial-analysis | xhigh effort | 33.5 % | independent | Artificial Analysis |
language
| Release 2026-06-25 tasks_counted=3 · livebench_version=2026-06-25 | 73.7 % | independent | LiveBench |
long context instruction
| AA-LCR | xhigh effort | 75.7 % | independent | Artificial Analysis |
| IFBench | xhigh effort | 77.6 % | independent | Artificial Analysis |
| Release 2026-06-25 tasks_counted=4 · livebench_version=2026-06-25 | 66.5 % | independent | LiveBench |
reasoning math
| Release 2026-06-25 tasks_counted=4 · livebench_version=2026-06-25 | 88.8 % | independent | LiveBench | |
| Release 2026-06-25 tasks_counted=4 · livebench_version=2026-06-25 | 77.7 % | independent | LiveBench |
Agent + model results
systems, not bare-model scores
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Verified | 72.8 % | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench bash-only | 72.8 % | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench bash-only | 72.8 % | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Multilingual | 66.3 % | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Multilingual | 66.3 % | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Multilingual | 0.7 usd_per_task | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Multilingual | 0.7 usd_per_task | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench Verified | 0.4 usd_per_task | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench bash-only | 0.4 usd_per_task | unverified | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + GPT 5.2 Codex | SWE-bench bash-only | 0.4 usd_per_task | unverified | SWE-bench Leaderboard |
| agent + model deepagent-harbor + GPT 5.2 Codex | Terminal-Bench 2.0 | 66.5 % | unverified | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + GPT 5.2 Codex | Terminal-Bench Hard | 37.1 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.