DeepSeek: DeepSeek V3

Text input Text output
Author's Description

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](https://github.com/deepseek-ai/DeepSeek-V3) for more information, or see the [launch announcement](https://api-docs.deepseek.com/news/news1226).

Key Specifications
Cost
$$$
Context
163K
Released
Dec 26, 2024
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Stop Top P Seed Min P Frequency Penalty Response Format Max Tokens Presence Penalty Temperature
Features

This model supports the following features:

Response Format
Performance Summary

DeepSeek-V3, created on December 26, 2024, demonstrates competitive response times, performing among the faster models with a 53rd percentile ranking in speed across six benchmarks. It also offers competitive pricing, ranking in the 58th percentile for cost efficiency. Notably, DeepSeek-V3 exhibits exceptional reliability, achieving a perfect 100th percentile, indicating minimal technical failures and consistent, usable responses. The model excels in several key areas. It achieved perfect accuracy in the Ethics benchmark, making it the most accurate model at its price point and among models of similar speed. Its General Knowledge performance is also strong at 99.5% accuracy, placing it in the 84th percentile. DeepSeek-V3 shows robust instruction following capabilities with 71.0% accuracy (89th percentile) and strong reasoning abilities at 84.0% accuracy (84th percentile). While its Coding performance is solid at 89.0% accuracy (78th percentile), its duration for this benchmark is in the lower half of models. Email Classification is another strong suit, with 98.0% accuracy. Overall, DeepSeek-V3 stands out for its high accuracy in critical domains, particularly ethics and general knowledge, coupled with outstanding reliability.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.2
Completion $0.8

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
DeepInfra
DeepInfra | deepseek/deepseek-chat-v3 163K $0.2 / 1M tokens $0.8 / 1M tokens
Novita
Novita | deepseek/deepseek-chat-v3 64K $0.4 / 1M tokens $1.3 / 1M tokens
Nebius
Nebius | deepseek/deepseek-chat-v3 163K $0.5 / 1M tokens $1.5 / 1M tokens
Fireworks
Fireworks | deepseek/deepseek-chat-v3 131K $0.9 / 1M tokens $0.9 / 1M tokens
DeepInfra
DeepInfra | deepseek/deepseek-chat-v3 163K $0.38 / 1M tokens $0.89 / 1M tokens
Targon
Targon | deepseek/deepseek-chat-v3 163K $0.3 / 1M tokens $0.85 / 1M tokens
Chutes
Chutes | deepseek/deepseek-chat-v3 163K $0.2 / 1M tokens $0.8 / 1M tokens
SiliconFlow
SiliconFlow | deepseek/deepseek-chat-v3 163K $0.2 / 1M tokens $0.8 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Free Executions Accuracy Cost Duration
Other Models by deepseek