GPT-5.4 Pro
OpenAI·
LLMsreasoningcloud
Overview
Highest-capability model optimized for quality and depth over speed.
Capabilities and innovations
Highest capabilityDecision-ready outputsExtended reasoningNative computer use1M token context (API)ARC-AGI-2 Leader (83.3%)Quality over Speed Optimization
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 94.4% | third party GPQA Diamond 94.4 — GPT-5.4 Pro benchmark table via AI Release Tracker, which transcribes OpenAI's reported set (BrowseComp 89.3, FrontierMath, GPQA 94.4, GDPval 82). Aggregator (lowest source tier); value corroborated by multiple secondary sources but not on independent boards (benchlm/vals.ai don't list GPT-5.4 Pro). Sits above the benchlm GPQA-D leader (Gemini 3.1 Pro 94.3) but is not contradicted by it. |
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: not reported by the vendor
- MMLU: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: not reported by the vendor
- MMMU-Pro: not reported by the vendor
Reliability
- Hallucination rate (Vectara HHEM)
- 8.3% (lower is better)independent
- Agentic tool use (τ-bench)
- 80.1% (higher is better)third party
Links
More from OpenAI
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.