Best cell per row highlighted. Null means undisclosed — never counted as zero.
| Provider | Alibaba Cloud | Mistral AI | Meta |
| Family / tier | Qwen3 · Large | Mistral Small · Small | Llama 4 · Maverick |
| Status | GA | GA | GA |
| Released | 2025-04-29 | 2026-03-16 | 2025-04-05 |
| Knowledge cutoff | 2024-10 | 2026-01 | 2024-08 |
| Context window | 131K | 256K | 1M |
| Max output tokens | 33K | 16K | 8K |
| Input modalities | text | text, image | text, image |
| Reasoning mode | optional | optional | none |
| Open weights | Yes | Yes | Yes |
| Input | $0.2 | $0.15 | $0.2 |
| Output | $0.6 | $0.6 | $0.85 |
| Cached input | — | — | — |
| Batch input | — | $0.075 | — |
| Free tier | Yes | Yes | No |
| Output speed (tok/s) | 68 | 180 | 104.3 |
| Time to first token (s) | 0.78 | — | 0.66 |
| Latency tier | medium | fast | fast |
| Trains on inputs | No | No | No |
| License | Apache-2.0 | Apache-2.0 | Llama-4-Community |
| Clouds | GCP | Bedrock, Azure AI Foundry, Vertex AI | Bedrock, Vertex AI, Azure AI Foundry, GCP, OCI, IBM watsonx |
| Compliance | — | SOC2, ISO27001, GDPR | — |
| Unified score | 8.5 | 8.5 | 7.7 |
| Decision Maker | 8 | 8.5 | 8.5 |
| Domain Strategist | 8.5 | 8.5 | 7.5 |
| Finance Lead | 9.5 | 9.5 | 9 |
| Domain Practitioner | 9 | 9 | 7.5 |
| Power User | 7.5 | 7.5 | 6.5 |
| Skeptic | 7 | 7.5 | 5.5 |
| Value score | 9.5 | 9.5 | 9 |
| MMLU | — | — | 85.5 |
| MMLU-Pro | 82.8 | 78 | 80.5 |
| GPQA Diamond | 70 | 71.2 | 69.8 |
| AIME 2025 | 81.5 | — | — |
| MATH-500 | — | — | 61.2 |
| HumanEval | — | — | 85.8 |
| LiveCodeBench | 70.7 | — | 43.4 |
| Aider Polyglot | — | — | 15.6 |
| MMMU | — | — | 73.4 |
| LMArena Elo | 1423 | — | 1327 |
| LMArena Coding Elo | 1472 | — | 1373 |
Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.