Best cell per row highlighted. Null means undisclosed — never counted as zero.
| Provider | Anthropic | OpenAI | |
| Family / tier | Gemini 3 · Pro | Claude 4 · Opus | GPT-5 · Pro |
| Status | preview | GA | GA |
| Released | 2026-02-19 | 2026-05-28 | 2026-04-24 |
| Knowledge cutoff | 2025-01 | 2026-01 | 2025-12 |
| Context window | 1.0M | 1M | 1.1M |
| Max output tokens | 66K | 128K | 128K |
| Input modalities | text, image, audio, video | text, image | text, image |
| Reasoning mode | optional | always | optional |
| Open weights | No | No | No |
| Input | $2 | $5 | $5 |
| Output | $12 | $25 | $30 |
| Cached input | $0.2 | $0.5 | $0.5 |
| Batch input | $1 | $2.5 | $2.5 |
| Free tier | No | No | No |
| Output speed (tok/s) | 142.7 | 60 | 68.7 |
| Time to first token (s) | 30.78 | 0.5 | 64 |
| Latency tier | slow | medium | slow |
| Trains on inputs | No | No | No |
| License | Proprietary | Proprietary | Proprietary |
| Clouds | Vertex AI, GCP | Bedrock, Vertex AI, Azure AI Foundry | Azure OpenAI, Azure AI Foundry |
| Compliance | SOC2, HIPAA, GDPR, ISO27001, FedRAMP, CCPA | SOC2, HIPAA, GDPR, ISO27001 | SOC2, GDPR, CCPA, HIPAA |
| Unified score | 8.7 | 9.2 | 8.7 |
| Decision Maker | 9 | 9 | 9 |
| Domain Strategist | 9 | 9 | 8.8 |
| Finance Lead | 8.5 | 8.5 | 7 |
| Domain Practitioner | 8.5 | 9.5 | 9.2 |
| Power User | 8.5 | 9 | 9 |
| Skeptic | 7.5 | 8 | 7.8 |
| Value score | 8.5 | 8 | 6.5 |
| MMLU | 92.6 | — | — |
| MMLU-Pro | — | — | 92.4 |
| GPQA Diamond | 94.3 | 93.6 | 93.6 |
| SWE-bench Verified | 80.6 | 88.6 | 88.7 |
| LiveCodeBench | 2887 | — | — |
| MMMU | 80.5 | — | 81.2 |
| TAU-bench | 90.8 | — | 98 |
| Terminal-Bench | 68.5 | 74.6 | 82.7 |
| MRCR Long Context | 84.9 | — | 74 |
| Humanity's Last Exam | 44.4 | — | 41.4 |
| LMArena Elo | 1487 | 1481 | 1482 |
| LMArena Coding Elo | 1520 | 1533 | 1520 |
Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.