DeepSeek Updates Online Model to V3.1-Terminus With Improved Stability

Breaking NewsSep. 23, 2025

Summarized by NextFin AI

DeepSeek has upgraded its V3.1 model to DeepSeek-V3.1-Terminus, enhancing core functionalities while addressing user-reported issues.
The new version improves stability in outputs, particularly in the Code Agent and Search Agent functions.
All platforms, including the app, web, mini program, and API, have been updated to the latest version, ensuring consistent user experience.

AsianFin -- DeepSeek announced on Monday that its V3.1 model has been upgraded to DeepSeek-V3.1-Terminus.

The latest version retains the core capabilities of the previous release while addressing user-reported issues, including mixed Chinese-English output and occasional abnormal characters. It also delivers enhanced performance in the Code Agent and Search Agent functions.

According to the company, the new version offers more stable outputs compared with V3.1. DeepSeek’s official app, web platform, mini program, and API have all been updated to DeepSeek-V3.1-Terminus.

Explore more exclusive insights at nextfin.ai.

Insights

What are the core capabilities of DeepSeek's V3.1 model?

How did user feedback influence the development of DeepSeek-V3.1-Terminus?

What specific issues were addressed in the latest update of DeepSeek?

What improvements were made in the Code Agent and Search Agent functions?

How does DeepSeek-V3.1-Terminus compare to its previous versions?

What impact does the new version have on user experience?

What platforms have been updated to DeepSeek-V3.1-Terminus?

Are there any notable trends in the online model development industry?

What are the potential future developments for DeepSeek's technology?

How might the improvements in DeepSeek affect its competitors?

What challenges does DeepSeek face in maintaining model stability?

Have there been any controversies surrounding DeepSeek’s updates?

What is the significance of stability in online model outputs?

How do mixed-language outputs impact user interactions with AI models?

What are some historical examples of similar updates in AI technology?

How does DeepSeek's performance compare to other AI language models?

What are the long-term implications of improved functionality in AI models?

NextFin.Al

No Noise, only Signal.

Open App