Yi-Large
01.AI · 2024-05-13
Yi-Large is 01.AI's closed, API-hosted text language model, globally released on May 13, 2024 as the company's largest model at the time. 01.AI documents it for chat and text generation, complex inference, prediction, natural-language understanding, and multilingual use. The API documentation labels its context window only as 32K, and 01.AI has not disclosed an exact parameter count.
Benchmark scores
| Benchmark | Score |
|---|---|
| BigCodeBench Complete | 47.2 |
| BigCodeBench Instruct | 37.7 |
| BigCodeBench-Hard Complete | 19.6 |
| BigCodeBench-Hard Instruct | 15.5 |
| Capability | 118.6 |
| MMLU | 79.3 |
| ZeroEval CRUX | 60.4 |
| ZeroEval GSM8K | 82.6 |
| ZeroEval MMLU-Redux | 82.2 |
| ZeroEval ZebraLogic | 18.9 |