| DeepSeek-R1 671B MoE | 3.99 | Coding assistance, logic-based Q&A, advanced reasoning, and chat |
| |
| DeepSeek-V3-0324 | 0.80 | Enhanced style and content quality |
| |
| Llama-4-Maverick-17B-128E-Instruct | 0.20 | Natively multimodal AI enables text and multimodal experiences |
|
|
| Llama-4-Scout-17B-16E-Instruct | 0.10 | More powerful than all previous generation Llama models |
|
|
| Llama-3.2-11B-Vision-Instruct | 0.15 | Object detection and description in images and video |
|
|
| Llama-3.2-3B-Instruct | 0.06 | Faster multi-language inference and translation |
|
|
| Llama-3.3-70B-Instruct | 0.50 | General Q&A and Retrieval Augmented Generation |
|
|
| Phi-4-mini-instruct | 0.12 | Fast, high-quality reasoning |
| |
| QwQ-32B | 0.65 | Enhanced thinking and reasoning performance. Fast coding. | | |
| Qwen2.5-VL-7B-Instruct | 0.15 | Object detection and description in images and video |
| |