Rankings / OpenAI
GPT 5.3 Codex
released 2026-02-05
BenchAtlas Index
as of 2026-07-14
xhigh effort
rank #29
8 families · 4 categories · medium
Benchmark evidence
10 results
agentic coding
| τ²-Bench | xhigh effort | 86.0 % | independent | Artificial Analysis |
coding
| SciCode | xhigh effort | 53.2 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | xhigh effort | 44.3 points | independent | Artificial Analysis |
| ECI | 155.5 points | independent | Epoch AI Benchmarking Hub | |
| ECI | xhigh effort | 155.5 points | independent | Epoch AI Benchmarking Hub |
| ECI | high effort | 155.5 points | independent | Epoch AI Benchmarking Hub |
knowledge science
| GPQA Diamond implementation=artificial-analysis | xhigh effort | 91.5 % | independent | Artificial Analysis |
| Humanity's Last Exam implementation=artificial-analysis | xhigh effort | 39.9 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | xhigh effort | 74.0 % | independent | Artificial Analysis |
| IFBench | xhigh effort | 75.4 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model Epoch Inspect harness + GPT 5.3 Codex | SWE-bench Verified | 74.8 % | independent | Epoch AI Benchmarking Hub |
| agent + model sage + GPT 5.3 Codex | Terminal-Bench 2.0 | 78.4 % | unverified | Terminal-Bench Leaderboard |
| agent + model Droid + GPT 5.3 Codex | Terminal-Bench 2.0 | 77.3 % | unverified | Terminal-Bench Leaderboard |
| agent + model ruley + GPT 5.3 Codex | Terminal-Bench 2.0 | 75.8 % | unverified | Terminal-Bench Leaderboard |
| agent + model codelia + GPT 5.3 Codex | Terminal-Bench 2.0 | 75.7 % | unverified | Terminal-Bench Leaderboard |
| agent + model simple_codex + GPT 5.3 Codex | Terminal-Bench 2.0 | 75.1 % | community | Terminal-Bench Leaderboard |
| agent + model mux + GPT 5.3 Codex | Terminal-Bench 2.0 | 74.6 % | unverified | Terminal-Bench Leaderboard |
| agent + model spoox-m + GPT 5.3 Codex | Terminal-Bench 2.0 | 71.5 % | unverified | Terminal-Bench Leaderboard |
| agent + model IndusAGICodingAgent + GPT 5.3 Codex | Terminal-Bench 2.0 | 69.1 % | unverified | Terminal-Bench Leaderboard |
| agent + model terminus-2 + GPT 5.3 Codex | Terminal-Bench 2.0 | 64.7 % | community | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + GPT 5.3 Codex | Terminal-Bench Hard | 53.0 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.