Best cell per row highlighted. Null means undisclosed — never counted as zero.
| Provider | OpenAI | OpenAI | OpenAI |
| Family / tier | GPT-5 · Pro | GPT-6 · Pro | GPT-6 · Pro |
| Status | GA | GA | GA |
| Released | 2026-04-24 | 2026-09-29 | 2026-09-22 |
| Knowledge cutoff | 2025-12 | 2026-04 | 2026-04 |
| Context window | 1.1M | 1.1M | 1.1M |
| Max output tokens | 128K | 128K | 128K |
| Input modalities | text, image | text, image | text, image |
| Reasoning mode | optional | always | optional |
| Open weights | No | No | No |
| Input | $5 | $2 | $2 |
| Output | $30 | $10 | $10 |
| Cached input | $0.5 | $0.1 | $0.2 |
| Batch input | $2.5 | $1 | $1 |
| Free tier | No | No | No |
| Output speed (tok/s) | 68.7 | 53 | — |
| Time to first token (s) | 64 | 5.5 | — |
| Latency tier | slow | medium | medium |
| Trains on inputs | No | No | No |
| License | Proprietary | Proprietary | Proprietary |
| Clouds | Azure OpenAI, Azure AI Foundry | Azure OpenAI, Azure AI Foundry, Bedrock | Azure OpenAI, Azure AI Foundry, Bedrock |
| Compliance | SOC2, GDPR, CCPA, HIPAA | SOC2, ISO27001, GDPR, CCPA, HIPAA | SOC2, ISO27001, GDPR, CCPA, HIPAA |
| Unified score | 8.7 | 9.5 | 9 |
| Decision Maker | 9 | 9.4 | 8 |
| Domain Strategist | 8.8 | 9 | 7.5 |
| Finance Lead | 7 | 9.5 | 8.8 |
| Domain Practitioner | 9.2 | 9.2 | 9 |
| Power User | 9 | 8.8 | 8.2 |
| Skeptic | 7.8 | 8 | 7.4 |
| Value score | 6.5 | 9.5 | 9 |
| MMLU-Pro | 92.4 | — | — |
| GPQA Diamond | 93.6 | 95.4 | 94.3 |
| SWE-bench Verified | 88.7 | — | — |
| MMMU | 81.2 | — | — |
| TAU-bench | 98 | — | — |
| Terminal-Bench | 82.7 | — | — |
| MRCR Long Context | 74 | — | — |
| Humanity's Last Exam | 41.4 | 52.9 | 47.9 |
| LMArena Elo | 1482 | — | 1457 |
| LMArena Coding Elo | 1520 | — | 1520 |
| Artificial Analysis Index | — | 52 | 48 |
Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.