Bytedance's Seed2.0 adds even more price pressure on Western AI models
Bytedance has released its new Seed2.0 AI model series. The story is a familiar one by now: the models match Western AI models on benchmarks while costing a fraction of the price.
The series comes in three sizes—Pro, Lite, and Mini—plus a dedicated code model. Bytedance says multimodal processing got the biggest upgrade in Seed2.0: the models are designed to better understand documents, tables, graphics, and videos. According to Bytedance, Seed2.0 Pro hits top scores on most benchmarks for visual math, logic, and perception, beating GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro in several categories.
The top-performing Seed2.0 Pro also reached gold medal-level scores in international math and programming competitions, including the International Mathematical Olympiad (IMO), the Chinese Mathematical Olympiad (CMO), and the international programming competition ICPC. Bytedance says Seed2.0 Pro earned gold in all five ICPC competitions tested, outperforming both Gemini 3 Pro and GPT-5.2. On the IMO, the model scored 35 out of 42 points but failed problem 6, just like every other model.
Seed2.0 still trails Claude in code generation and falls behind Gemini in long-tail knowledge, Bytedance says. The model also underperforms Western competitors when it comes to avoiding hallucinations, according to Bytedance's own benchmarks.
| Benchmark | GPT-5.2 High | Claude-Sonnet-4.5 | Claude-Opus-4.5 | Gemini-3-pro High | Seed2.0 Pro |
|---|---|---|---|---|---|
| Terminal Bench 2.0¹ | 62.4 | 45.2 | 60.2 | 56.9 | 55.8 |
| SWE-Lancer | 48.9 | 45.7 | 56.1 | 44.3 | 49.4 |
| SWE Bench Verified | 80.0 | 77.2 | 80.9 | 76.2 | 76.5 |
| Multi-SWE-Bench | 47.7 | 47.7 | 52.8 | 50.2 | 45.2 |
| SWE-Bench Pro | 55.6 | 48.4 | 55.4 | 49.7 | 46.9 |
Pricing undercuts Western models by a wide margin
The Seed2.0 Pro API costs roughly $0.47 per million input tokens and $2.37 per million output tokens. For comparison, Anthropic's Claude Opus 4.5—currently one of the most capable Western AI models—starts at $5.00 per million input tokens and $25.00 per million output tokens.
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| GPT-5.2 High | $1,75 | $14,00 |
| Claude Opus 4.5 Thinking | $5,00 | $25,00 |
| Gemini 3 Pro | $2,00-4,00* | $12,00-18,00* |
| Seed2.0 Pro | $0,47 (¥3,41) | $2,37 (¥17,04) |
| Seed2.0 Lite | $0,09 (¥0,64) | $0,53 (¥3,83) |
| Seed2.0 Mini | $0,03 (¥0,22) | $0,31 (¥2,24) |
*Gemini prices vary depending on context length and input modality according to Google's pricing page.
Seed2.0 Pro is available through the Chinese chat app Doubao, while the code model can be accessed through the developer tool TRAE. The API for all Seed2.0 variants runs on Bytedance's cloud platform Volcano Engine. The project page with technical details is currently only available in Chinese. Additional information is available in the model card.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe now