Term
DeepSeek V4
DeepSeek V4 (April 2026) is the fourth generation of the Chinese open-weight MoE model — released as Pro (1.6T parameters) and Flash (284B) under MIT license, with a 1M-token context window.
DeepSeek V4 — explained in more detail
DeepSeek V4 was released as a preview on April 24, 2026 and is the fourth generation of the DeepSeek model family. The architecture builds on Mixture-of-Experts with a dual-mode design (Thinking / Non-Thinking) and a 1 million token context window. Weights are released under MIT license — V4-Pro is now the largest open-weight model on the market, ahead of Kimi K2.6 (1.1T) and GLM-5.1 (754B).
Variants and pricing
- DeepSeek-V4-Pro: 1.6 trillion parameters total / 49 billion active per token. API pricing: $1.74 per million input tokens, $3.48 per million output tokens.
- DeepSeek-V4-Flash: 284B total / 13B active. API pricing: $0.14 input / $0.28 output per million tokens — significantly undercutting GPT-5.4 Nano and Claude Sonnet 4.6.
Where it fits
According to DeepSeek’s own paper, V4-Pro trails the closed frontier models by roughly 3 to 6 months, though it can partially close that gap through expanded reasoning tokens. Efficiency gains over V3.2: only 27% of the single-token FLOPs and 10% of the KV cache size at 1M-token context. The official API model names deepseek-chat and deepseek-reasoner will be deprecated on July 24, 2026.
Entdecke mehr
Three Days with Claude Fable 5 — and Why Opus Stays My Daily Driver
Fable 5 fixed a bug in one shot that Opus failed at twice. I'm still not switching. My honest impressions from the trenches after three days.
GlossarDeepSeek Janus-Pro-7B
Janus-Pro-7B is DeepSeek's open-source multimodal model from January 2025 — unifying image understanding and image generation in a single 7B architecture, MIT-licensed.
LexikonComparing AI models — who builds what and how to choose
The major model families in 2026 at a glance. Who builds Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen — and which model to pick when.