Skip to content
LIVE
Loading prices...
DeepSeek Releases New Thinking Model to Challenge U.S. 

A picture of two robots confronting each other

DeepSeek Releases New Thinking Model to Challenge U.S. 

In Brief

  • • DeepSeek has launched DeepSeek-V3.2 and its Speciale variant, to compete with top U.S. models.
  • • The new models integrate enhanced reasoning and tool-use capabilities based on an earlier model.
  • • These developments intensify the ongoing AI race between China and the U.S

The AI race has gotten tougher as Chinese company DeepSeek rolls out a new thinking-first AI model. Known as DeepSeek-V3.2, the model also has a variant said to be on par with OpenAI’s GPT-5 and Google’s Gemini-3 Pro.

Ad

In an announcement on 2 December, the company said it has also released DeepSeek-V3.2 Speciale with superior reasoning capabilities. This is part of DeepSeek’s strategy to get ahead in the race to develop thinking AI models to compete in the evolving space.

DeepSeek-V3.2 replaces Deepseek-V3.2-Exp, an experimental version that was launched in September as a test version prior to the release of this final version

DeepSeek Reaching New Levels of Competence

DeepSeek took the AI industry by storm in January this year after it launched a free model that gives users access to all its services at no cost. While the model launched with a lot of promise, it also faced a lot of challenges such as legal hurdles in many countries which hindered the company from making progress.

The company also faced allegations of user privacy violations in several countries, leading to its several bans. Now, it seems DeepSeek is set to pick up the challenge and step up its game. DeepSeek-V3.2 is the first model it has built that has thinking integrated directly into tool-use, and also supports tool-use in both thinking and non-thinking modes.

Ad

DeepSeek-V3.2 Speciale is another big addition to DeepSeek’s arsenal. Designed specially for long thinking, the model can handle the most difficult tasks such as solving mathematical problems and algorithmic reasoning, performing at near-gold level in the International Math Olympiad and other elite contests which sets it at a higher level than most open-source models.

DeepSeek Models Make AI Race Tougher for The U.S.

The race between the U.S. and China in terms of AI development has been on for a long time now. It started with U.S. companies like OpenAI creating ChatGPT and Google creating Gemini, while Chinese companies like Alibaba have Qwen 3 and DeepSeek now has DeepSeek-v3.2.

With the new models released by DeepSeek and Alibaba releasing its own AI assistant, China may be getting ahead of the U.S. in AI development as suggested by Nvidia CEO Jensen Huang.

More Must-Reads:

How do you rate this article?

Join our Socials

Briefly, clearly and without noise – get the most important crypto news and market insights first.