Model providers
20 providers we read; 11 have at least one documented fact.
| Provider | Model versions | With documented facts |
|---|---|---|
| Anthropic | 21 | 21 |
| DeepSeek | 10 | 2 |
| Google DeepMind | 50 | 22 |
| InclusionAI (Ant Group) | 18 | 1 |
| Meta | 8 | 4 |
| MiniMax | 7 | 1 |
| Moonshot AI | 4 | 1 |
| OpenAI | 33 | 18 |
| Thinking Machines Lab | 3 | 2 |
| xAI | 4 | 4 |
| Z.ai (Zhipu AI) | 7 | 2 |
No document we read states a capability level, a test or a safeguard for any model version of: Alibaba Qwen (12 versions), Dots Studio (1 version), Mistral AI (4 versions), NVIDIA (39 versions), poolside (4 versions), StepFun (8 versions), Tencent Hunyuan (16 versions), Upstage (3 versions), Xiaomi MiMo (8 versions).