Highlights
- DeepSeek-V3.1-Terminus update is now live on Android, iOS, web, and API.
- The update improves language consistency and fixes multilingual output issues, while upgrading Code and Search Agents.
- The model shows improved scores across key benchmarks like MMLU-Pro, GPQA-Diamond, and SWE Verified.
Caption – DeepSeek V3.1-Terminus Update Rolls Out. (Image credit – DeepSeek)
DeepSeek has officially released its latest update, DeepSeek-V3.1-Terminus. It builds on the DeepSeek V3 model family first launched in December 2024. This new release is an upgraded version of the DeepSeek-V3.1 model introduced two months ago and is designed to resolve issues highlighted by users while improving overall performance. Here’s what we know.
🚀 DeepSeek-V3.1 → DeepSeek-V3.1-Terminus
The latest update builds on V3.1’s strengths while addressing key user feedback.✨ What’s improved?
🌐 Language consistency: fewer CN/EN mix-ups & no more random chars.
🤖 Agent upgrades: stronger Code Agent & Search Agent performance.…— DeepSeek (@deepseek_ai) September 22, 2025
DeepSeek-V3.1-Terminus – Key Improvements
The update brings significant enhancements in language consistency and agent upgrades, directly addressing problems noted in user feedback. According to DeepSeek, earlier versions occasionally produced mix-ups between Chinese and English text along with abnormal character outputs. These issues have now been fixed in the V3.1-Terminus version.
The company has also upgraded its Code Agent and Search Agent, strengthening DeepSeek’s task-specific frameworks for better usability. Furthermore, the platform continues to offer two distinct operational modes –
- deepseek-chat (non-thinking mode)
- deepseek-reasoner (thinking mode)
These refinements aim to deliver more stable, reliable, and accurate outputs across use cases.
Benchmark Performance Gains
The improvements are reflected in benchmark scores, where DeepSeek-V3.1-Terminus outperforms its predecessor. Compared to DeepSeek-V3.1, the new version achieved higher results in multiple categories that are as follows –
- MMLU-Pro: 85.0 (up from 84.8)
- GPQA-Diamond: 80.7 (up from 80.1)
- Humanity’s Last Exam: 21.7 (up from 15.9)
- LiveCodeBench: 74.9 (up from 74.8)
- BrowseComp: 38.5 (up from 30.0)
- SimpleQA: 96.8 (up from 93.4)
- SWE Verified: 68.4 (up from 66.0)
- SWE-bench Multilingual: 57.8 (up from 54.5)
- Terminal-bench: 36.7 (up from 31.3)
📊 DeepSeek-V3.1-Terminus delivers more stable & reliable outputs across benchmarks compared to the previous version.
👉 Available now on: App / Web / API
🔗 Open-source weights here: https://t.co/Jh4RudofKmThanks to everyone for your feedback. It drives us to keep improving… pic.twitter.com/6fdLvl4LG3
— DeepSeek (@deepseek_ai) September 22, 2025
These gains highlight the update’s stronger reasoning abilities, coding performance, and multilingual handling.
Availability of DeepSeek-V3.1-Terminus
The new update is now available across multiple platforms including Android, iOS, web, and API. It is also live on Hugging Face with integration into AnyCoder on Hugging Face and NovitaLabs serverless API underway.
FAQs
Q1. What is the DeepSeek-V3.1-Terminus update?
Answer. It’s the latest upgrade to DeepSeek’s V3 model family, focused on fixing multilingual output issues and enhancing overall performance and stability.
Q2. What improvements does the DeepSeek-V3.1-Terminus update bring?
Answer. It improves language consistency, upgrades the Code and Search Agents, and boosts benchmark scores across reasoning, coding, and multilingual tasks.
Q3. Where is DeepSeek-V3.1-Terminus available?
Answer. The update is live on Android, iOS, web, and API platforms, with integrations underway on Hugging Face and NovitaLabs’ serverless API.