DeepSeek R1 Distill Llama 70B

Creator

DeepSeek

Released

2025-01-20

Intelligence

9.9

Artificial Analysis Index

Coding

Artificial Analysis Index

In $/1M

$0.70

input tokens

Out $/1M

$1.05

output tokens

Blended $/1M

$0.79

3:1 blended

Speed

0

tokens / sec

Profile

License, openness, and modality — the governance layer.

License

MIT

commercial OK

Weights

Open

downloadable

Modalities

Text

Parameters

71B

total

DeepSeek-R1 is the first-generation reasoning model built atop DeepSeek-V3 (671B total parameters, 37B activated per token). It incorporates large-scale reinforcement learning (RL) to enhance its chain-of-thought and reasoning capabilities, delivering strong performance in math, code, and multi-step reasoning tasks.

Capability profile

Category strength across 20 domains, via LLM Stats.

math
90general
80reasoning
80biology
70
physics
70chemistry
70code
60
Via LLM Stats

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
40
Humanity’s Last Exam
6
MMLU-Pro
80
SciCode
31
LiveCodeBench
27
MATH-500
93
AIME 2025
54
τ²-Bench (agentic)
22
Terminal-Bench Hard
2
IFBench
28
Via Artificial Analysis
© 2026 NYSGPT2525 LLC