Archive · 3 Aug 2026
DeepSeek’s V4‑Flash: Cheapest AI Model Worldwide
DeepSeek is a Chinese company that makes computer programs that can answer questions and write text.
They just launched a new version called V4‑Flash.
It is the cheapest one to run, costing only a few cents for a test.
It still works well, scoring the same as some big companies’ models.
DeepSeek plans a stronger version called V4‑Pro, but it isn’t ready yet.
DeepSeek released V4‑Flash, the lowest‑cost AI model on benchmark tests.
V4‑Flash runs at $0.14 per million input tokens and $0.28 per million output tokens.
Artificial Analysis estimates average cost of 3 cents per test, over 100 times cheaper than Anthropic’s Claude Fable 5.
The model scores 50/100 on the Intelligence Index, matching Google’s Gemini 3.6 Flash.
DeepSeek plans a higher‑performance V4‑Pro version, release date unknown.
- Who
- DeepSeek, a Chinese AI startup
- What
- released its V4‑Flash AI model, the cheapest on benchmark tests
- When
- released on Friday (2025)
- Where
- Beijing, China
- Why
- to offer ultra‑low‑cost AI alternatives and regain market momentum
- Model
- V4‑Flash
- Input token cost
- $0.14 per million
- Output token cost
- $0.28 per million
- Average cost per test
- $0.03
- Intelligence Index score
- 50/100
