DeepSeekOpen SourceCodingReasoning
DeepSeek V4
DeepSeek’s V4 generation — deepseek-v4-pro (1.6T / 49B active) and deepseek-v4-flash (284B / 13B active) with 1M context, dual thinking modes, and strong agentic coding. Flash-0731 is the current Flash API revision.
Tool calling · Thinking · Coding
Last reviewed: 31 July 2026
Overview
DeepSeek’s V4 generation — deepseek-v4-pro (1.6T / 49B active) and deepseek-v4-flash (284B / 13B active) with 1M context, dual thinking modes, and strong agentic coding. Flash-0731 is the current Flash API revision.
Capabilities
- Vision: No
- Audio: No
- Tool calling: Yes
- Thinking: Yes
- MCP: No
- Coding: Yes
- Structured output: Yes
Technical specifications
- Provider
- DeepSeek
- License
- Model License (open weights; verify per release)
- Context window
- 1M
- Parameters
- Pro MoE 1.6T/49B active; Flash MoE 284B/13B active
- Architecture
- Mixture-of-Experts (DSA / sparse attention)
- Release
- 2026-04
- Modalities
- Text
- Vision
- No
- Audio
- No
- Tool calling
- Yes
- Thinking
- Yes
- MCP
- No
- Open weights
- Yes
- API
- Yes
- Pricing (input)
- DeepSeek API V4 Flash/Pro rates (verify platform)
- Pricing (output)
- DeepSeek API V4 Flash/Pro rates (verify platform)
Supported modalities
Text
Context window
1M (1,000,000 tokens)
Pricing
Input: DeepSeek API V4 Flash/Pro rates (verify platform)
Output: DeepSeek API V4 Flash/Pro rates (verify platform)
Use model IDs deepseek-v4-pro and deepseek-v4-flash. Older deepseek-chat / deepseek-reasoner aliases retired Jul 24 2026.
Availability
API: Yes
Chat UI: Yes
Open weights: Yes
API at api.deepseek.com (OpenAI/Anthropic-compatible); open weights per DeepSeek V4 release notes.
Use cases
- Cost-efficient agentic coding
- 1M-context document and repo work
- OpenAI-compatible / Codex-style agent backends
- Self-hosted open-weight deployments
Strengths
- 1M context as default across V4 services
- Strong Flash agent upgrades (0731) at low activated params
- Open weights plus cheap API economics
Limitations
- Text-first vs frontier multimodal VLMs
- Enterprise packaging thinner than OpenAI/Anthropic
- Pro vs Flash quality/latency tradeoffs need workload evals
Related guides
Related benchmarks
Related research
- DeepSeek V4 Preview Release
Original Paper
Related GitHub
Related tools
Related rankings
Companies
Explore more models
- DeepSeek R1DeepSeekDeepSeek’s reasoning-focused model trained with reinforcement learning for multi-step math, science, and coding problem solving.
- DeepSeek V3DeepSeekDeepSeek’s MoE general model — strong open-weight performance on coding and knowledge tasks with competitive API pricing.
- Kimi K3Moonshot AIMoonshot’s Kimi K3 — 2.8T MoE (104B active) open-weight multimodal agentic model with 1M context, native vision, and strong long-horizon coding. Weights on Hugging Face under the Kimi K3 License.
- Qwen3AlibabaAlibaba’s Qwen3 generation — strong multilingual open models spanning chat, reasoning modes, coding, and multimodal variants.
- Llama 4MetaMeta’s Llama 4 family — open-weight multimodal models designed for research and commercial use under Meta’s community license.
- PhiMicrosoftMicrosoft’s Phi family of small language models — high capability per parameter for on-device, edge, and cost-sensitive deployments.