Seed-1.6: benchmark results
ByteDance's Doubao-Seed-1.6 (June 11, 2025), a 230-billion-parameter, 23-billion-active MoE foundation model with Adaptive Chain-of-Thought and integrated text, image and short-video understanding in a 256K context. Provider: ByteDance. Released 2025-06-11. Access: API.
Aggregate Rating 1672 ± 7, rank #265 of 1656 rated models, from 335 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FIRE (Finance, Qualification Exams) - Banking (CCBP) | 92.16 | Accuracy (%) on the Banking (CCBP) questions of FIRE's finan | 100 |
| FIRE (Finance, Qualification Exams) - Economist | 90.64 | Accuracy (%) on the Economist (Jr/Int/Sr) questions of FIRE' | 100 |
| FIRE (Finance, Qualification Exams) - Futures Practitioner | 91.9 | Accuracy (%) on the Futures Practitioner questions of FIRE's | 100 |
| FIRE (Finance, Qualification Exams) - Public Accountant (CPA) | 93.12 | Accuracy (%) on the Public Accountant (CPA) questions of FIR | 100 |
| FinToolBench - Tool Execution Success Rate | 32.54 | Tool Execution Success Rate (TESR, 0-1, times 100): share of | 100 |
| GameplayQA - Object Recognition | 77.1 | Accuracy (%) on the 70 Object Recognition questions (the mod | 100 |
| LabOSBench | 76.3 | Subtask Success Rate (%; subtask-averaged per instrument, ma | 100 |
| LabOSBench - Activation & Alignment | 59.1 | Subtask Success Rate (%; mean over the category's operation- | 100 |
| LabOSBench - Experimental Execution | 82.1 | Subtask Success Rate (%; mean over the category's operation- | 100 |
| LabOSBench - Post-processing & Completion | 81.2 | Subtask Success Rate (%; mean over the category's operation- | 100 |
| LabOSBench - Preparation & Sample Handling | 97.1 | Subtask Success Rate (%; mean over the category's operation- | 100 |
| MT-AgentRisk - Filesystem | 78.57 | Attack success rate (%): share of the 70 Filesystem-MCP task | 100 |
Interactive version: theaggregate.ai/model?slug=seed-1-6 · How It Works · Data refreshed daily, snapshot 2026-10-20.