LLM ReferenceLLM Reference

DeepSeek R1

Open Source

About

DeepSeek R1: Reasoning-optimized model with extended thinking capabilities. 128K context.

Capabilities

VisionMultimodalReasoningFunction CallingTool UseStructured OutputsCode Execution

Providers(12)

Compare all →
ProviderInput (per 1M)Output (per 1M)Type
DeepSeek Platform$0.55$2.19Serverless
OpenRouter$0.7$2.5Serverless
Together AI$3.0$7.0Serverless
Fireworks AI$0.56$1.68Serverless
NVIDIA NIMServerless
Microsoft FoundryServerless
Databricks Foundation Model ServingServerless
DeepInfraServerless
AWS BedrockServerless
Bitdeer AI$0.1$0.3Serverless
SiliconFlow$0.25$0.8Serverless
Replicate API$3.75$10Serverless

Benchmark Scores(5)

BenchmarkScoreVersionSource
HumanEval89.92025-01DeepSeek
SWE-bench Verified49.22025-01DeepSeek
Aider Polyglot56.92026-04https://aider.chat/docs/leaderboards
Chatbot Arena1372.0https://lmarena.ai
Google-Proof Q&A71.5diamondhttps://arxiv.org/abs/2501.12948

Rankings

Specifications

Released2025-01-20
Parameters671B, 37B Active
Context128K
ArchitectureDecoder Only
Specializationgeneral
Trainingmultistage
Fine-tuningtask_specific

Created by

Advancing artificial general intelligence (AGI).

Hangzhou, Zhejiang, China
Founded 2023
Website