DeepSeek, a Chinese AI company, has launched two advanced reasoning models – V3.2 and V3.2-Speciale – that claim to outperform leading US models like Gemini 3.0 Pro and GPT-5 High on key benchmarks. While the standard V3.2 is available to users through their app and web interface, the more powerful V3.2-Speciale is currently accessible only via API.
New Models Push Boundaries in AI Reasoning
V3.2 is an upgrade from DeepSeek’s earlier experimental version, V3.2-Exp, released in September. The company asserts that V3.2 matches Gemini 3.0 Pro in reasoning performance and exceeds OpenAI’s GPT-5 in certain tests. Testing against GPT-5 Pro was not provided, which may be because ChatGPT is blocked in China.
Crucially, DeepSeek maintains an open-source approach, allowing free access to its models for any organization. However, running V3.2 requires substantial computing power: with 685 billion parameters, it demands a large server infrastructure with significant hardware investments.
The Rise of Chinese AI Competition
DeepSeek’s release is significant because it directly challenges the dominance of American AI giants. The company first gained recognition earlier this year by releasing a high-performing reasoning model for free, demonstrating that advanced AI capabilities can be achieved at lower costs.
This move has pressured OpenAI, Google, and Anthropic to accelerate their own development of reasoning-level models, with all three now offering free or subscription-based access to comparable technologies. The competitive landscape is shifting.
Academic Success and Future Implications
DeepSeek-V3.2-Speciale has notably achieved top scores in prestigious international competitions: winning gold at the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). This demonstrates the model’s advanced problem-solving abilities.
DeepSeek’s entrance into the market is a clear indication that AI innovation is no longer solely concentrated in the US. The open-source nature of their models, combined with their proven performance, could force American companies to rethink their strategies and prioritize faster development cycles.
This is a developing situation, and further independent verification of DeepSeek’s claims is needed. However, the company has already changed the narrative around AI reasoning and forced the industry to re-evaluate its approach to innovation.
