Overview
DeepSeek's fast general-purpose text model with 1M context and JSON mode
Key Features
- 284B total parameters with 13B activated per token via Mixture-of-Experts
- Hybrid Compressed Sparse Attention for efficient 1M token context handling
- Manifold-Constrained Hyper-Connections for improved training stability
- Text-only model — supports code, documents, and structured data inputs
- Fast, cost-effective — ideal for high-throughput chat and coding tasks
Input Capabilities
Text
Available Tools
JSON Mode
Output responses in valid JSON format.
Technical Specifications
Context Window
1,000,000
Max Output
64,000
Model Information
Provider: DeepSeek
Model Code: deepseek-v4-flash
Category: Chat
File Support
Max Files
10
Max Size
10MB
Supported Formats
.txt
.md
.csv
.json
.html
.css
+10 more