DeepSeek's latest V4.1-Flash-1017 AI model is faster and leaner
What's the story
Chinese tech firm DeepSeek has launched an updated version of its artificial intelligence (AI) model, dubbed the V4.1-Flash-1017. The company claims that the new model offers "stronger benchmark performance, faster task completion, and shorter outputs than the previous version." The update is now available across the web, on DeepSeek's app and its platform for developers.
Performance boost
New model scores better on Humanity's Last Exam
The new model from DeepSeek, a minor update to its previous V4.1-Flash version, is said to be better at reasoning, coding, and visual tasks.
It scored up to 42.6 on Humanity's Last Exam (HLE), an improvement over the previous version's score of 39.1.
HLE tests how well AI models answer thousands of questions across math, sciences, and humanities topics.
Versatility
DeepSeek claims substantial improvements across various tasks
The Hangzhou-based start-up said in its technical report on Hugging Face, an online platform for AI models, datasets, and apps, that the new model brings "substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks."
These enhancements make it suitable for a wide range of applications from game creation to 3D modeling.
Technical difficulties
DeepSeek recently faced technical difficulties
Notably, DeepSeek has been facing some technical difficulties. Earlier this week, the company suffered an outage of about five hours, leaving its servers unreachable.
Some experts have attributed the glitch to insufficient computing resources.
The issues have led to slowdowns and service disruptions in recent months for DeepSeek, whose competitively priced AI models are in high demand among developers.