Baichuan 2 (7B and 13B)
Open 7B and 13B bilingual models trained from scratch on 2.6T tokens, matching or beating similar-size open models, strong in medicine and law.
Followed Baichuan-7B (2023-06-13) and Baichuan-13B (2023-07-08). The paper (arXiv 2309.10305, 2023-09-19) promises all pretraining checkpoints.
- Date
- Wednesday, 6 September 2023
- Lab
- Baichuan
- Kind
- open-weights
- Access
- open weights (restricted license)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Pretraining tokens | 2.6T | company |
Release day 2023-09-06 is from memory; HF repos created 2023-08-29 UTC and the paper is 2023-09-19. Treat the day as unconfirmed.
Sources
This record was partly confirmed: some claims could not be checked on 6 October 2026. How we check