This is their latest experimental model, and it’s packed with upgrades 👇
⚡ Built on V3.1-Terminus for faster, more efficient training & inference on long context.
🧠 DeepSeek Sparse Attention (DSA) makes it leaner and smarter for big workloads.
💻 Available now in both the app and API.
💰 API prices cut by 50%+ — massive cost drop for devs!
Plus:
📉 Input: $0.028 / Output: $0.42
🧾 Docs & comparisons available on their official site.
It’s a clear signal — efficiency and affordability are becoming the new standard in open AI models.
🔥 Open source, fast context, low cost… the trifecta.
👉 What do you think — will DeepSeek become a serious alternative in your stack?