Doubao Seed 2.0 Pro
ByteDance's 'agent era' flagship — GPT-5.2-class reasoning at roughly a tenth of the price.
About
Doubao Seed 2.0 Pro is the flagship of ByteDance's Seed 2.0 series, released February 14, 2026 (Reuters-confirmed) alongside Lite, Mini, and Code variants. ByteDance positions it for the 'agent era' — executing complex real-world tasks rather than just answering questions — with the Pro version carrying complex reasoning and multi-step task execution the company says matches OpenAI's GPT-5.2 and Google's Gemini 3 Pro at roughly an order of magnitude lower cost. The Doubao app itself counts 155M weekly active users in China.
The benchmark sheet is provider-reported: AIME 2025 98.3, Codeforces rating 3020, SWE-Bench Verified 76.5%, GPQA Diamond 88.9, VideoMME 89.5, plus competition-level ICPC/IMO/CMO golds. Honest caveats from third-party reviews: it does not beat GPT-5.4 on coding benchmarks, and it does not match Claude Opus 4.7 (87.6% on SWE-Bench Verified). At $0.47 input / $2.37 output per million tokens on Volcano Engine, it is 5–10x cheaper than Western frontier models while delivering 85–90% of their benchmark ceiling. Context window is 256K tokens.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported: AIME 2025 98.3, Codeforces 3020, SWE-Bench Verified 76.5%, GPQA Diamond 88.9, VideoMME 89.5.
Independent: third-party comparisons confirm it trails GPT-5.4 on coding and Claude Opus 4.7 (87.6%) on SWE-Bench Verified.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Release confirmed by Reuters with consistent specs across independent outlets; pricing and benchmark claims corroborated by third-party reviews that also publish the honest caveats. Docked for: (1) benchmark figures are provider-run; (2) proprietary with no weights; (3) international access leans on third-party routing.