DeepSeek Ships V4-Flash, Beats Its Own Pro Model
The new release reportedly outscores V4-Pro's preview and rival GLM-5.2 on benchmarks, with a full Pro release next.
What's happening
DeepSeek released the official V4-Flash model, and multiple outlets including ithome, cnbeta and Nikkei report its agent-capability benchmarks exceed the V4-Pro preview edition and Zhipu's GLM-5.2. It scored 50 on the Artificial Analysis Intelligence Index and 76.7 on Cybergem for a 200B-parameter model, and 36kr reports a full V4-Pro release is coming soon.
The numbers
Velocity 6.0 is the highest in this batch, breadth 3, 23 recent items including a 329-point Hacker News thread, concentrated almost entirely on the model release itself rather than tangential chatter. No price data applies since DeepSeek isn't a traded asset here.
Our read
This reads as a real technical release — benchmark claims are specific and repeated across independent outlets rather than vague hype. Watch for independent benchmark verification outside DeepSeek's own claims and whether the promised V4-Pro full release actually ships.
Sources we checked
- DeepSeek-V4-Flash Update — Hacker News
- DeepSeek-V4-Flash正式版 launched, taking on the Pro version — cnbeta.com.tw
- DeepSeek-V4-Flash Official Version API Goes Online for Public Testing, Agent Capability Benchmark Tests Far Exceed V4-Pro Preview Edition — ithome.com
- DeepSeek releases beta version of V4 models as AI price war heats up — asia.nikkei.com
- DeepSeek-V4-Flash正式版 launches public testing with performance surpassing Zhi谱GLM-5.2 — cnbeta.com.tw
- DeepSeek: DeepSeek-V4-Pro Official Version Will Be Released Soon — 36kr.com
- #DeepSeek #V4Flash 0731, a new AI model, scores 50 on the Artificial Analysis Intelligence Index, surpassing its predecessor and placing it just behind GPT-5.6 — @technews@eicker.news
- DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis — Hacker News
See today's live board
This story is frozen as published on August 1, 2026. The live board shows what is moving right now, scored against real money.
Open this story live → See today's live boardEarlier on DeepSeek
More AI stories
- Alibaba's Qwen 3.8-Max Lands With Real Weight August 4, 2026
- Qwen3.8-Max goes open-weight at 2.4 trillion parameters August 3, 2026
- DeepSeek cuts prices, pauses fundraising August 2, 2026
- Altman Declares Singularity Arrived, Heads to Washington on Policy July 29, 2026
- Amodei Says Anthropic Does Not Support Open-Weight AI Ban July 28, 2026
- Sam Altman Heads to the White House With Singularity Claims in Tow July 28, 2026
Research, not financial advice. Sort Brick reports what its data showed at the time of publication. It is not a recommendation to buy or sell anything — you decide.