OpenAI/GPT-5/Feb 5, 2026
GPT-5.3 Codex
OpenAI's most capable agentic coding model. The first to combine Codex + GPT-5 training stacks, bringing together best-in-class code generation, reasoning, and general-purpose intelligence. Sets new state-of-the-art on Terminal-Bench 2.0 (77.3%) and OSWorld-Verified (64.7%). ~25% faster than GPT-5.2.
textcodereasoningvisiontool-useaudio
Arena ELO
—
Input Price
$1.75/M
Output Price
$14/M
Speed
73 t/s
Context
400K
Latency
125010ms
Capability Assessment
SWE-Bench Pro80.0%
Terminal-Bench 2.077.3%
GPQA Diamond73.8%
MMMU Pro84.0%
Comparative Analysis
| Metric | GPT-5.3 Codex | Claude Opus 4.6 | Gemini 3.1 Pro | Grok 4 |
|---|---|---|---|---|
| SWE-bench | 80.0% | 80.8% | 80.6% | 72.0% |
| AIME 2025 | 94.0% | 100.0% | 91.2% | 94.0% |
| GPQA Diamond | 73.8% | 91.3% | 94.3% | 88.0% |
| MMLU | — | 91.0% | 92.6% | — |
| Input $/M | $1.75 | $5 | $2 | $3 |
| Output $/M | $14 | $25 | $12 | $15 |