Best cell per row highlighted. Null means undisclosed — never counted as zero.
| Provider | Anthropic | Anthropic |
| Family / tier | Claude 4 · Opus | Claude 4 · Opus |
| Status | GA | GA |
| Released | 2025-11-23 | 2026-05-27 |
| Knowledge cutoff | 2025-05 | — |
| Context window | 200K | 1M |
| Max output tokens | 64K | 128K |
| Input modalities | text, image | text, image |
| Reasoning mode | optional | always |
| Open weights | No | No |
| Input | $5 | $5 |
| Output | $25 | $25 |
| Cached input | $0.5 | $0.5 |
| Batch input | $2.5 | $2.5 |
| Free tier | No | No |
| Output speed (tok/s) | 53.2 | 60 |
| Time to first token (s) | 1.58 | 0.5 |
| Latency tier | slow | medium |
| Trains on inputs | No | No |
| License | Proprietary | Proprietary |
| Clouds | Bedrock, Vertex AI, Azure AI Foundry | Bedrock, Vertex AI, Azure AI Foundry |
| Compliance | SOC2, HIPAA, GDPR, ISO27001 | SOC2, HIPAA, GDPR, ISO27001 |
| Unified score | 8 | 9.2 |
| Decision Maker | 7.5 | 9 |
| Domain Strategist | 8 | 9 |
| Finance Lead | 8 | 8.5 |
| Domain Practitioner | 8 | 9.5 |
| Power User | 8 | 9 |
| Skeptic | 7.5 | 8 |
| Value score | 7.5 | 8 |
| MMLU | 90.8 | — |
| MMLU-Pro | 90 | — |
| GPQA Diamond | 87 | 93.6 |
| SWE-bench Verified | 80.9 | 88.6 |
| HumanEval | 92 | — |
| MMMU | 80.7 | — |
| TAU-bench | 88.9 | — |
| Terminal-Bench | 59.8 | 74.6 |
| Humanity's Last Exam | 43.4 | — |
| LMArena Elo | 1468 | — |
| LMArena Coding Elo | 1510 | — |
| Artificial Analysis Index | 43 | — |
Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.