DeepSeekOpen SourceDec 26, 2024

Deepseek Chat

DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions. Pre-trained on nearly 15 trillion tokens, the reported evaluations reveal that the model outperforms other open-source models and rivals leading closed-source models. For model details, please visit the DeepSeek-V3 repo for more information, or see the launch announcement.

Context Window
164K
tokens
Max Output
164K
tokens
Released
Dec 26, 2024
Arena Rank
Output Speed
175
tokens/sec
Time to First Token
650ms
TTFT

Capabilities

👁Vision
🧠Reasoning
🔧Tool Calling
Prompt Caching
🖥Computer Use
🎨Image Generation

Supported Parameters

Frequency Penalty
Reduce repetition
Max Tokens
Output length limit
min_p
Presence Penalty
Encourage new topics
Repetition Penalty
Penalize repeated tokens
Response Format
JSON mode / structured output
Seed
Deterministic outputs
Stop Sequences
Custom stop tokens
structured_outputs
Temperature
Controls randomness
Tool Choice
Control tool usage
Tool Calling
Function calling support
Top K
Top-K sampling
Top P
Nucleus sampling

Pricing Comparison

RouterInput / 1MOutput / 1MCached Input / 1M
Requesty$0.28$0.42$0.03
OpenRouter$0.30$1.20$0.15
Martian$0.30$1.20$0.15

Benchmarks

Artificial Analysis
Intelligence IndexArtificial Analysis
74/100
Coding IndexArtificial Analysis
70/100
Math IndexArtificial Analysis
82/100
MMLU-PROArtificial Analysis
0.815/1
GPQA DiamondArtificial Analysis
0.758/1
MATH-500Artificial Analysis
0.942/1
AIME 2024Artificial Analysis
0.68/1
Humanity's Last ExamArtificial Analysis
0.1/1
LiveCodeBenchArtificial Analysis
0.695/1
SciCodeArtificial Analysis
0.38/1

Model IDs

Requestydeepseek/deepseek-chat
OpenRouterdeepseek/deepseek-chat

Tags

tool-callingcaching
Compare with another model

Related Models