How much does R1 Distill Llama 70B cost?

R1 Distill Llama 70B costs $0.80 per million input tokens and $0.80 per million output tokens. For a typical conversation (~2,000 tokens), that's approximately $0.002 per message.

What benchmarks has R1 Distill Llama 70B been tested on?

R1 Distill Llama 70B has been evaluated on 8 benchmarks. Top scores: IFEval: 43.4, MMLU-PRO: 41.6, BBH (HuggingFace): 35.8.

Is R1 Distill Llama 70B open source?

Yes, R1 Distill Llama 70B is open source.

How does R1 Distill Llama 70B compare to Grok-3 mini?

R1 Distill Llama 70B has an average score of 47.6 while Grok-3 mini scores 47.7. Grok-3 mini slightly outperforms R1 Distill Llama 70B overall. See full comparison →

Home/Models/R1 Distill Llama 70B

R1 Distill Llama 70B

Name: R1 Distill Llama 70B
Price: 0.8 USD
Author: DeepSeek

by DeepSeek · Released Jan 2025

Open Source

47.6

avg score

Rank #143

Compare

Better than 48% of all models

Context

128K tokens (~64 books)

Input $/1M

$0.80

Output $/1M

$0.80

Type

text

License

Open Source

Benchmarks

8 tested

Data updated today

About

DeepSeek R1 Distill Llama 70B is a distilled large language model based on [Llama-3.3-70B-Instruct](/meta-llama/llama-3.3-70b-instruct), using outputs from [DeepSeek R1](/deepseek/deepseek-r1). The model combines advanced distillation techniques to achieve high performance across...

Tested on 8 benchmarks with 27.8% average. Top scores: IFEval (43.4%), MMLU-PRO (41.6%), BBH (HuggingFace) (35.8%).

Looking for similar performance at lower cost?
Nemotron 3 Ultra scores 48.1 (101% as good) at $0.50/1M input · 38% cheaper

Capabilities

reasoning

13.3

#143 globally

math

30.7

#156 globally

knowledge

21.8

#231 globally

speed

22.8

#77 globally

general

35.8

#25 globally

language

43.4

#128 globally

Benchmark Scores

Compare All

Tested on 8 benchmarks · Ranked across 6 categories

Score Distribution (all 274 models)

0255075100

▲ You are here

reasoningCompare reasoning →

MUSR

HuggingFace MuSR (Multi-Step Reasoning). Tests multi-hop reasoning requiring chaining multiple facts together.

13.3—

mathCompare math →

MATH Level 5

HuggingFace evaluation of MATH Level 5 problems. Competition math requiring advanced reasoning and proof construction.

30.7—

knowledgeCompare knowledge →

MMLU-PRO

HuggingFace MMLU-Pro. Harder version of MMLU with 10 answer choices instead of 4 and more challenging questions.

41.6—

GPQA

HuggingFace evaluation of GPQA (Graduate-Level Google-Proof Q&A). PhD-level science questions that cannot be easily searched.

2.0—

Quick compare:

vs Grok-3 mini

vs Nemotron 3 Ultra

vs Qwen3 4B Thinking 2507

Excellent (85+) Good (70-85) Average (50-70) Below (<50)

Similar Models

Qwen3 4B Thinking 2507

Alibaba

48.4TBD

Links

Info

DeepSeek Pricing explorer Developers · API

Research

Documentation

Community

Source Code

BenchGecko API

deepseek-r1-distill-llama-70b

Specifications

Typetext
Context128K tokens (~64 books)
ReleasedJan 2025
LicenseOpen Source
StatusActive
Cost / Message~$0.002

Available On

DeepSeek$0.80

Frequently Asked Questions

R1 Distill Llama 70B is an open-source text AI model by DeepSeek, released in January 2025. It has an average benchmark score of 47.6. Context window: 128K tokens.

Benchmarks

IFEval MMLU-PRO BBH (HuggingFace)MATH Level 5 Artificial Analysis — Quality Index

DeepSeek · Provider DeepSeek · Economy All Models Compare Models Pricing Developers · API Context Window · Glossary

R1 Distill Llama 70B

Frequently Asked Questions

Related Models

Benchmarks

Related Pages