DeepSeek

DeepSeek R1 Distill Llama 8B

Zero-eval

by DeepSeek

+
+
+
+
About

DeepSeek-R1-Distill-Llama-8B was developed as a compact distilled variant, designed to bring reasoning capabilities to a more efficient 8B parameter Llama base. Built to democratize access to reasoning-enhanced models, it provides a lightweight option for applications requiring analytical depth with limited resources.

+
+
+
+
Timeline
AnnouncedJan 20, 2025
ReleasedJan 20, 2025
+
+
+
+
Specifications
Training Tokens14.8T
+
+
+
+
License & Family
License
MIT
Performance Overview
Performance metrics and category breakdown

Overall Performance

4 benchmarks
Average Score
64.4%
Best Score
89.1%
High Performers (80%+)
2
+
+
+
+
All Benchmark Results for DeepSeek R1 Distill Llama 8B
Complete list of benchmark scores with detailed information
MATH-500
text
0.89
89.1%
Self-reported
AIME 2024
text
0.80
80.0%
Self-reported
GPQA
text
0.49
49.0%
Self-reported
LiveCodeBench
text
0.40
39.6%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+