DeepSeek, a Hangzhou-based AI startup, has released V4.1-Flash-1017 on the web, its app, and its developer platform.

The company said the update improves reasoning, coding, agentic, and visual tasks while completing tasks faster and producing shorter outputs than the previous V4.1-Flash model.

DeepSeek described the release as a minor update to V4.1-Flash.

In a technical report on Hugging Face, the online platform for AI models, datasets, and apps, the startup said the model delivers substantial improvements across those tasks.

The update follows a roughly five-hour outage earlier this week that left DeepSeek’s servers unreachable.

According to the source report, slowdowns and service disruptions have increased in recent months. The company maintains a status page.

The launch comes as Chinese AI providers compete on affordability. Alibaba Cloud’s broader expansion includes more than 100 AI models launched in September 2024.

🔗 Source: Bloomberg

Recent DeepSeek developments