Alibaba Unveils Qwen3.8-Max Model—China’s Latest AI Challenger To OpenAI And Anthropic
The Chinese tech giant hailed its model as its “most capable.”
DeepSeek has unveiled DeepSeek-V4-Flash-0731, a smaller model that outperforms its larger flagship model. The company released the update as a public beta via its API, with open weights published on Hugging Face under the MIT license. Despite using the same architecture as the preview release, DeepSeek claims additional post-training is responsible for the performance gains.
The Flash version, with 284 billion total parameters, now exceeds the V4-Pro preview in several agent-focused benchmarks, including Terminal-Bench 2.1, DeepSWE, and Toolathlon-Verified. The MIT license allows organizations greater control over deployment and customization. Features like support for familiar API formats and speculative decoding with DeepSeek’s DSpark framework make serving more efficient.
This release demonstrates that companies are finding new ways to enhance model effectiveness through delivery methods rather than simply increasing size.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- DeepSeek’s smaller model just outperformed its own flagship thenewstack.io
