DeepSeek launches V4-Flash, the cheapest AI model

Chinese AI startup DeepSeek released its V4-Flash model on 31 July 2026. The model is priced at $0.14 per million input tokens and $0.28 per million output tokens, which research firm Artificial Analysis translates to an average cost of about 3 cents per benchmark test. This makes V4-Flash more than 100 times cheaper to run than Anthropic’s Claude Fable 5 and substantially cheaper than OpenAI’s GPT‑5.6 and other leading models.

V4-Flash uses a mixture‑of‑experts architecture with 284 billion parameters and a 1‑million‑token context window. In performance testing it scored 50 out of 100 on the Intelligence Index, matching Google’s Gemini 3.6 Flash and trailing top models such as OpenAI’s GPT‑5.6, Anthropic’s Claude Opus 5, and Moonshot AI’s Kimi K3. The launch is part of DeepSeek’s strategy to regain market momentum by offering ultra‑low‑cost AI, and the company is also preparing a more powerful V4‑Pro version and a potential IPO. The pricing pressure adds to the broader competition between Chinese AI firms—including Alibaba, Moonshot AI, MiniMax and Z AI—and U.S. providers such as OpenAI, Anthropic and Google.