Best cell per row highlighted. Null means undisclosed — never counted as zero.
| Provider | Anthropic | Alibaba Cloud | OpenAI |
| Family / tier | Claude 5 · Fable | Qwen2.5-Coder · Coder | GPT-5 · Medium |
| Status | GA | GA | GA |
| Released | 2026-06-09 | 2024-11-12 | 2026-03-05 |
| Knowledge cutoff | 2026-01 | 2024-08 | 2025-08 |
| Context window | 1M | 131K | 1.1M |
| Max output tokens | 128K | 8K | 128K |
| Input modalities | text, image, file | text | text, image |
| Reasoning mode | always | none | optional |
| Open weights | No | Yes | No |
| Input | $10 | $0.08 | $2.5 |
| Output | $50 | $0.24 | $15 |
| Cached input | $1 | — | $0.25 |
| Batch input | $5 | — | $1.25 |
| Free tier | No | Yes | No |
| Output speed (tok/s) | — | — | 86.3 |
| Time to first token (s) | — | — | 1.2 |
| Latency tier | slow | fast | medium |
| Trains on inputs | No | No | No |
| License | Proprietary | Apache-2.0 | Proprietary |
| Clouds | Bedrock, Vertex AI, Azure AI Foundry | GCP | Azure OpenAI, Azure AI Foundry |
| Compliance | SOC2, HIPAA, GDPR, ISO27001 | — | SOC2, GDPR, CCPA, HIPAA |
| Unified score | 9.5 | 8.4 | 8.7 |
| Decision Maker | 9.5 | 8.5 | 9 |
| Domain Strategist | 9.5 | 8 | 8.7 |
| Finance Lead | 7 | 9 | 9 |
| Domain Practitioner | 9.5 | 9.5 | 9 |
| Power User | 9 | 8.5 | 8.5 |
| Skeptic | 7.5 | 7.5 | 8 |
| Value score | 6.5 | 9.5 | 8.5 |
| GPQA Diamond | — | — | 92.8 |
| SWE-bench Verified | 95 | — | — |
| HumanEval | — | 92.7 | — |
| LiveCodeBench | — | 31.4 | — |
| MMMU | — | — | 81.2 |
| Terminal-Bench | 88 | — | 75.1 |
| MRCR Long Context | — | — | 36.6 |
| Humanity's Last Exam | — | — | 39.8 |
| LMArena Elo | 1506 | 1270 | 1476 |
| LMArena Coding Elo | 1552 | 1342 | 1520 |
Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.