DeepSeek releases V4-Flash update
The release, and up to 50% lower API prices, followed a roughly $7.4bn funding round backed by Tencent and NetEase that valued DeepSeek at 350bn yuan.
- Models & capabilities
- Open weights & ecosystem
- Minor
DeepSeek released the official version of its V4-Flash model, Caixin Global reported, arriving slightly behind the company’s mid-July target and without the more capable V4-Pro variant that had also been expected. The architecture was reported as unchanged from an April preview version, with the update’s gains attributed entirely to further post-training rather than a larger or redesigned model; DeepSeek said the update brought stronger autonomous-agent capability and cut API prices by up to 50%.
The release came against a backdrop of intensifying competition among Chinese model developers on both capability and price: V4-Flash was reported to have been the most-used model on OpenRouter for seven consecutive weeks over the summer, but Caixin’s report said it still trailed Moonshot’s Kimi K3 on capability. DeepSeek had separately raised a funding round of roughly $7.4 billion backed by Tencent and NetEase, valuing the company at 350 billion yuan, with proceeds earmarked for doubling headcount and further agent development — as well as a previously announced, not-yet-implemented dynamic pricing scheme that would raise API costs during peak-demand hours.
The update was one of several within weeks from Chinese labs — alongside Moonshot’s Kimi K3 and Zhipu’s GLM — competing on price and agentic capability in a market where open-weight releases from China had become frequent enough that individual updates drew comparatively modest attention outside specialist coverage.