A new artificial intelligence (AI) model was spotted on AI gateway platform OpenRouter on March 11. It is believed that the model, named Hunter Alpha, may be DeepSeek’s next-generation system. DeepSeek is a Chinese firm that went viral last year after its model DeepSeek-R1 outperformed the likes of ChatGPT and Google Gemini.
Update: Xiaomi’s AI model team MiMo has confirmed that Hunter Alpha is an “early internal test build of MiMo-V2-Pro.” The model is designed to act as the “brain” of AI agents that will be able to complete tasks on behalf of humans.
As per a Reuters report, the developer has intentionally not disclosed their identity. OpenRouter describes Hunter Alpha a “stealth model.”
What is Hunter Alpha?
Hunter Alpha is available as a free model. It is described as a 1-trillion-parameter model, placing it among the most advanced AI systems currently available. It also features a context window of up to one million tokens, which allows it to process and retain vast amounts of information in a single interaction.
As per the report, during tests, the chatbot identified itself as a Chinese AI system primarily trained in Chinese, with knowledge updated until May 2025, matching the knowledge cutoff reported for DeepSeek’s existing models.
When asked about its creator, the system declined to provide any information, stating it only knew its name, parameter scale, and context window length.
When India Today Tech asked Hunter Alpha about its creators, the chatbot replied, “I was created by a group of engineers passionate about AGI, though I don’t have any specific details about who they are or their affiliations. I have 1T parameters and a 1M token context window.”
The chabot did not directly mention that it was built in China. However, it stated, “I’m designed to comply with the laws and regulations of the People’s Republic of China.”
Why Hunter Alpha could be a big deal
The report states that such a combination of scale, reasoning ability, and free access is unusual, as models with similar capabilities typically require significant computational resources. This suggests that it was likely created by a major firm from China, which might be DeepSeek.
The specifications of Hunter Alpha closely resemble expectations for DeepSeek’s upcoming V4 system, which Chinese media reports suggest could launch as early as April.
Since its release, Hunter Alpha has processed more than 160 billion tokens, driven largely by software development tools and AI agent frameworks.
Do note that DeepSeek and OpenRouter have not commented on such speculation.
In the past DeepSeek has managed to compete against models from OpenAI and Google at a fraction of the cost. For instance, DeepSeek-V3 is said to have cost less than $6 million in development, while OpenAI’s GPT-4 model cost over $100 million. The Chinese company is also facing accusations from Anthropic of potentially using Claude to train its models.
