{"id":196,"date":"2025-08-27T04:43:42","date_gmt":"2025-08-27T08:43:42","guid":{"rendered":"https:\/\/www.aiai.com\/blog\/?p=196"},"modified":"2025-09-04T06:35:08","modified_gmt":"2025-09-04T10:35:08","slug":"deepseek-v3-1-release","status":"publish","type":"post","link":"https:\/\/www.aiai.com\/blog\/deepseek-v3-1-release","title":{"rendered":"DeepSeek V3.1 Release: Higher Thinking Efficiency, the First Step Towards the Agent Era"},"content":{"rendered":"\n
\n
\n
<\/div>\n\n\n\n
\n

DeepSeek V3.1 Release: Higher Thinking Efficiency, the First Step Towards the Agent Era<\/strong><\/h1>\n\n\n\n

The new DeepSeek-V3.1 has launched, almost six months after the DeepSeek-V3-0324 model. There was no prior announcement.<\/p>\n\n\n\n

\"\"<\/figure>\n<\/div>\n\n\n\n
<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
\n
<\/div>\n\n\n\n

TABLE OF CONTENTS<\/strong><\/a><\/p>\n\n\n\n

What Does Hybrid Inference Bring?<\/a><\/p>\n\n\n\n

How Strong Are the Agent Capabilities?<\/a><\/p>\n\n\n\n

How Was the Technical Foundation Upgraded?<\/a><\/p>\n\n\n\n

Product Form and API System<\/a><\/p>\n\n\n\n

Community Experience and Discussion<\/a><\/p>\n\n\n\n

Practical Tips for Developers<\/a><\/p>\n\n\n\n

Conclusion<\/a><\/p>\n<\/div>\n<\/div>\n\n\n\n

\n

Since launching the DeepSeek-R1 model in January, DeepSeek has quietly made small updates to its models. First, the DeepSeek_V3_0324 model update arrived in March. Then, the DeepSeek-R1-0528 update came at the end of May. Both updates improved functionality and performance, but these were small changes. They weren’t big enough to need a new version number. The DeepSeek-V3.1 model is here. It shows a big boost in performance and competes strongly with OpenAI’s GPT-5 model.<\/p>\n\n\n\n

The core change can be summarized in one sentence: One model, two experiences. You can now use hybrid inference. This lets users easily switch between thinking mode and non-thinking mode with the DeepThink button. Switch to thinking mode for detailed reasoning. Use non-thinking mode for simple answers. Official end-to-end products have all been upgraded at once. The website, app, mini-program, and API now use the new model called DeepSeek-V3.<\/p>\n\n\n\n

\n
Try DeepSeek V3.1<\/a><\/div>\n<\/div>\n\n\n\n

Quick Summary<\/strong><\/p>\n\n\n\n

1. Hybrid inference mode, one model supporting free switching between thinking and non-thinking modes.<\/p>\n\n\n\n

2. Significant efficiency improvement with faster speed and approximately 20-50 fewer output tokens.<\/p>\n\n\n\n

3. Enhanced Agent capabilities with noticeably improved performance in programming and search agents.<\/p>\n\n\n\n

4. Support for 128K long context, further strengthened compatibility, and continued open-source commitment.<\/p>\n\n\n\n

What Does Hybrid Inference Bring?<\/h2>\n\n\n\n

V3.1-Think delivers results like R1-0528 on various benchmarks, but it does so more quickly. On AIME 2025, R1-0528 scored 87.5 while V3.1 achieved 88.4. On GPQA, R1-0528 scored 81 compared to V3.1’s 80.1. On liveCodeBench, R1-0528 reached 73.3 while V3.1 hit 74.8. While maintaining comparable quality, V3.1-Think reduces output tokens by approximately 20-50. We’ve achieved the goal of making the model both perform better and consume fewer resources.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Independent analysis provides additional perspective. Artificial Analysis notes that V3.1 combines V3 and R1 into a hybrid inference model. It scores 60 in reasoning mode, up from R1’s 59. In non-reasoning mode, it scores 49, a big jump from V3 0324’s 44.<\/p>\n\n\n\n

How Strong Are the Agent Capabilities?<\/h2>\n\n\n\n

DeepSeek says that V3.1 has made big strides in tool use and multi-step agent tasks. This post-training optimisation marks progress toward the Agent era.<\/p>\n\n\n\n

The team used an internal code agent framework to test code repair skills on the SWE-bench test. The results showed that fewer rounds were needed compared to the open-source OpenHands framework. Code repair abilities also improved significantly over earlier DeepSeek models.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Testing on Terminal-Bench with the official Terminus 1 framework showed that terminal agents handled complex command-line tasks more reliably.<\/p>\n\n\n\n

V3.1 offers a special search tool format for thinking mode. It supports complex multi-round search agents. In tests using commercial search APIs, web filters, and 128K context (with R1-0528 as a comparison), V3.1 showed clear benefits. It excelled in browsecomp tasks needing multi-step reasoning. There were also significant gains in the expert-level HLE challenge (text subset only). It handled complex problems well, especially those needing current or external information.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\n
Try DeepSeek V3.1<\/a><\/div>\n<\/div>\n\n\n\n

How Was the Technical Foundation Upgraded?<\/h2>\n\n\n\n

V3.1 is built on the brand-new V3.1-Base. The base model uses a two-stage method to extend context. It includes extensive training on the V3 checkpoint. The additional training volume amounts to 840 billion tokens.<\/p>\n\n\n\n

The 32K extension stage achieved a 10x boost in training volume, hitting 630 billion tokens.<\/p>\n\n\n\n

The 128K extension stage increased by 3.3x, reaching 209 billion tokens.<\/p>\n\n\n\n

The team significantly expanded the dataset scale for both stages by collecting more long documents.<\/p>\n\n\n\n

Training uses the UE8M0 FP8 scaling format and updates the tokenizer. Related settings are synced with the official GitHub and HuggingFace repositories. The team stated on the official WeChat public account that UE8MO FP8 is designed for the next generation of domestic chips. Developers planning to deploy should carefully read the documentation to avoid potential issues.<\/p>\n\n\n\n

Product Form and API System<\/h2>\n\n\n\n

One clear change is that the web interface now just labels everything as “Deep Thinking.” It has removed the R1 designation from the Deep Thinking UI.” The online model has also expanded the context window from 64K to 128K. Additionally, DeepSeek has launched a new API service architecture:<\/p>\n\n\n\n