0

R1 Distill Llama 70B

DeepSeek R1 Distill Llama 70B is an AI model from DeepSeek.

Publisher
DeepSeek
Family
Deepseek
Context
131K
Input
Output
Released
20 Jan 2025

Cite

Notes

Only stored in your browser.

Attribution

Benchmark scores
AAOpenLLMLiveBench
Attribution policy →
Context
131K
Input
$0.70
/1M tok
Output
$1.05
/1M tok
Blended
$0.79
#287/ 414/1M · 3:1
Speed
47
#192/ 414tok/s

Price & speed via Artificial Analysis · ranked across tracked models

Reported on 19 evals across 7 domains - best 93.5% on MATH-500 · #35 of 178

Reported eval scores

19
AIME 2024: Problems from the American Invitational Mathematics Examination67
MATH-50093.5
MATH Level 530.7
LiveBench - Instruction Following70.2
LiveBench - Coding55.5
LiveBench - Language29.5
Terminal-Bench (Hard)1.5
IFBench27.6
MuSR43.4
BIG-Bench Hard (BBH)56.3
Humanity's Last Exam (HLE)6.1
LiveBench51.7
MMLU-Pro47.5
IFEval43.4
AIME 2025: Problems from the American Invitational Mathematics Examination53.7
SciCode31.3
GPQA Diamond26.5
LiveCodeBench26.6
τ²-bench (Tau²-bench)21.9