World Knowledge on MMLU-Pro (EM)
73.5EM (World Knowledge)DeepSeek-V4-Pro Base
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-V4-Pro Base# Shots=5-shot, Architecture=MoE, # Activated Params=49B, # Total Params=1.6T2026.04 | 73.5 | |
| DeepSeek-V4-Flash Base# Shots=5-shot, Architecture=MoE, # Activated Params=13B, # Total Params=284B2026.04 | 68.3 | |
| DeepSeek-V3.2 Base# Shots=5-shot, Architecture=MoE, # Activated Params=37B, # Total Params=671B2026.04 | 65.5 | |
| Sigma-MoE-Tiny Base# Shots=5-shot, Architecture=MoE, # Activated Params=0.5B, # Total Params=20B2025.12 | 38.13 | |
| Gemma-3 4B Base# Shots=5-shot, Architecture=Dense, # Activated Params=4B, # Total Params=4B2025.12 | 29.23 | |
| Qwen3 0.6B Base# Shots=5-shot, Architecture=Dense, # Activated Params=0.6B, # Total Params=0.6B2025.12 | 24.74 |