Seed 2.0 (Pro, Lite, Mini, Code)
Flagship Doubao generation in Pro, Lite, Mini and Code models; Pro claims IMO, CMO and ICPC gold results and token prices about ten times lower.
Tuned for production traffic and long-tail knowledge. Pro tops GPT-5.2 on SuperGPQA, and Seed ranked 6th on LMArena text and 3rd on vision (2026-02-16). Pro scores BrowseComp 77.3, SWE-Bench Pro 46.9 and Terminal Bench 2.0 55.8. Closed API.
- Date
- Saturday, 14 February 2026
- Lab
- ByteDance Seed
- Kind
- model
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| SuperGPQA (Pro) | 68.7% blog says this beats GPT-5.2 | company |
| BrowseComp (Pro) | 77.3% | company |
| SWE-Bench Pro (Pro) | 46.9% | company |
| Terminal Bench 2.0 (Pro) | 55.8% | company |
| LMArena text rank | 6th as of 2026-02-16, per Seed blog | company |
Date from the Seed blog (2026-02-14). All benchmark numbers are ByteDance's own; the 'order of magnitude cheaper' pricing claim is the company's. Weights are not released.
Sources
- seed.bytedance.com/en/blog/seed2-0-%E6%AD%A3%E5%BC%8F%E5%8F%91%E5%B8%83
- seed.bytedance.com/en/seed2
- technode.com/2026/02/14/bytedance-releases-doubao-seed-2-0-positions-pro-model-against-gpt
This record was checked against its sources on 6 October 2026. How we check