Share your thoughts, 1 month free Claude Pro on usSee more

Contextual Reasoning on HellaSwag

88EM

DeepSeek-V4-Pro Base

Updated 1mo ago

Evaluation Results

Method	Links
DeepSeek-V4-Pro Base 2026.04		88
DeepSeek-V3.2 Base 2026.04		86.4
DeepSeek-V4-Flash Base 2026.04		85.7
DeepSeek-V2 Lite 2025.12		80.55
Sigma-MoE-Tiny Base 2025.12		79.79
Gemma-3 4B Base 2025.12		77.76
Qwen3 0.6B Base 2025.12		53.57