Two Minute Papers · 2026-08-03 · notable
Two Minute Papers — 'Another DeepSeek Moment Has Arrived'
Károly Zsolnai-Fehér walks through DeepSeek V4-Flash 0731, the newly official Chinese model that hits 82.7 on Terminal-Bench 2.1 while running at a fraction of frontier-lab pricing.
Two Minute Papers frames DeepSeek V4-Flash 0731 as the next 'DeepSeek moment' — a low-cost Chinese model catching the frontier again.
What is it?
This Two Minute Papers episode covers DeepSeek V4-Flash 0731, the version DeepSeek promoted from April preview to official on July 31, and asks whether it triggers another cost shock like the original DeepSeek V3 did last year. Zsolnai-Fehér works through the model's coding scores, its price sheet, and where the sudden jump between the preview and the 0731 build came from.
How does it work?
The video keeps DeepSeek's architecture unchanged from the April preview and attributes the score jump — 82.7 on Terminal-Bench 2.1, close to Claude Opus 4.8 — entirely to extra post-training. Two Minute Papers' usual side-by-side coding demos show the model driving agent workflows that were previously in the frontier-lab-only tier.
Why does it matter?
DeepSeek V4-Flash 0731 is the reason 'race to zero' is trending again this week: a Chinese lab shipping near-frontier coding performance at a small fraction of frontier prices. This Two Minute Papers episode is the fastest visual explainer of what the 0731 build actually does, useful for teams weighing whether to move an agent workload off Opus or Sol.