DeepSeek efficient MoE with 128K context. 671B parameters. Exceptional code (94/100) and math (93/100). State-of-the-art performance at low inference cost.
Metrics
ECI Score145.113
Epoch Capabilities Index score
ECI Normalized92
Normalized ECI score (0-100)
Context Window128,000
Maximum input token capacity
Training Compute1.0e24 FLOP
Total compute used during training
Score Breakdown
code
20%100
Code generation, understanding, and debugging
math
20%100
Mathematical reasoning and problem solving
reasoning
15%99
Multi-step logical reasoning
intelligence
30%92
Overall reasoning and task completion ability
instruction following
96
Compatibility(7)
Sources(3)
Developed by
DeepSeekBenchmark Performance
Compare ModelsEpoch Capabilities Index
145.1/ 155
| Dimension | Score |
|---|---|
| Intelligence | 92.0 |
| Reasoning | 99.0 |
| Math | 100.0 |
| Code | 100.0 |
| Instruction Following | 96.0 |
| Overall | 92 |
Loading chart...
Data: Epoch AI (CC-BY 4.0)