GPT-5.3 CodexvsGPT-5.2-Codex
Across 8 shared benchmarks, GPT-5.3 Codex leads overall: GPT-5.3 Codex wins 6, GPT-5.2-Codex wins 2, with 0 ties and an average score difference of +3.43.
GPT-5.3 Codex
OpenAI · 2026-02-05 · Coding model
GPT-5.2-Codex
OpenAI · 2025-12-18 · Coding model
GPT-5.3 Codex6 wins(75%)(25%)2 winsGPT-5.2-Codex
Benchmark scores
Grouped by capability, sorted by largest gap within each. 8 shared benchmarks.
Agent Level Benchmark
Even 2/2| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| Terminal Bench Hard | 5317 / 244Extra-High (With Tools) | 37.1065 / 244Extra-High (With Tools) | +15.90 |
| τ²-Bench - Telecom | 8685 / 264Extra-High (With Tools) | 92.1054 / 264Extra-High (With Tools) | -6.10 |
General Knowledge
GPT-5.3 Codex 2/2| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| CritPt | 16.9047 / 200Extra-High (No Tools) | 8.7078 / 200Extra-High (No Tools) | +8.20 |
| HLE | 42.50109 / 563Extra-High (No Tools) · Text only | 35.70164 / 563Extra-High (No Tools) · Text only | +6.80 |
General Evaluation
GPT-5.3 Codex 1/1| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| GPQA Diamond | 91.5061 / 462Extra-High (No Tools) | 89.9086 / 462Extra-High (No Tools) | +1.60 |
Instruction Following
GPT-5.2-Codex 1/1| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| IF Bench | 75.4034 / 282Extra-High (No Tools) | 77.6016 / 282Extra-High (No Tools) | -2.20 |
Long Context
GPT-5.3 Codex 1/1| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| AA-LCR | 83.3011 / 170Extra-High (No Tools) | 82.3020 / 170Extra-High (No Tools) | +1 |
Multimodal Understanding
GPT-5.3 Codex 1/1| Benchmark | GPT-5.3 Codex | GPT-5.2-Codex | Diff |
|---|---|---|---|
| MMMU-Pro | 78.5059 / 227Extra-High (No Tools) | 76.3075 / 227Extra-High (No Tools) | +2.20 |
Specs
| Field | GPT-5.3 Codex | GPT-5.2-Codex |
|---|---|---|
| Publisher | OpenAI | OpenAI |
| Release date | 2026-02-05 | 2025-12-18 |
| Model type | Coding model | Coding model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 400K | Not available |
| Max output | 125K | Not available |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.3 Codex | GPT-5.2-Codex |
|---|---|---|
| Text input | $1.75 / 1M tokens | $1.25 / 1M tokens |
| Text output | $14 / 1M tokens | $10 / 1M tokens |
| Cache read | $0.175 / 1M tokens | $0.125 / 1M tokens |
| Cache write | Not public | $0 / 1M tokens |
Summary
- GPT-5.3 Codexleads in:General Knowledge (2/2), General Evaluation (1/1), Long Context (1/1), Multimodal Understanding (1/1)
- GPT-5.2-Codexleads in:Instruction Following (1/1)
- Tied in:Agent Level Benchmark
On average across the 8 shared benchmarks, GPT-5.3 Codex scores 3.43 higher.
Largest single-benchmark gap: Terminal Bench Hard — GPT-5.3 Codex 53 vs GPT-5.2-Codex 37.10 (+15.90).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.