Skip to content
DeepSeek

DeepSeek V4

Open Source

DeepSeekReleased on

DeepSeek's trillion-parameter MoE model with only 32B active parameters. 1M context window, native multimodal (text/image/video), and API pricing at ~1/20th of GPT-5. The fastest adopted open-source model in history, capturing 6% global market share within months.

85
Overall Score

Core Specs

1000K
Context Window
66K
Max Output
Reasoning
Open Source
Multimodal Support
textimagevideo

User Feedback Highlights

Based on community feedback. Hover to see original reviews.

+ 1M token context window Servers in China (latency for overseas users) Geopolitical supply chain concerns Self-hosting requires significant hardware+ Native multimodal (text/image/video)+ Price 1/20th of proprietary models+ Open-source & self-hostable+ Matches frontier model performance
Sentiment:👍 70%😐 25%👎 5%

Pros & Cons

Pros

  • +1M token context window
  • +Native multimodal (text/image/video)
  • +Price 1/20th of proprietary models
  • +Open-source & self-hostable
  • +Matches frontier model performance

Cons

  • Servers in China (latency for overseas users)
  • Geopolitical supply chain concerns
  • Self-hosting requires significant hardware

Reliability

SLA99.0%
Incidents (30d)1
Last Incident2026-03-30
Self-hosting available. API servers in China.
View Status Page →

Pricing

Input (per 1M tokens)$0.14
Output (per 1M tokens)$0.28
Free trial available
Updated on 2026-03-08

Get Started

1Visit the provider's website
2Create an account
3Start using the model

Benchmarks

sweBench%
mmlu91%
humanEval94%
arcAgi2%
noteMatches GPT-5 level on most benchmarks%