A new LLM chat product that can produce near-instant responses (for prompts that do not involve a web search, that is) is out.
This new feature apparently runs completely on non-NVIDIA hardware:
>On Thursday, Mistral released an app called Le Chat that it said can respond to user questions with 1,000 words per second.
>Cerebras said it is providing the computer power behind those results, which it claimed makes Mistral the world's fastest AI assistant, ahead of both OpenAI and DeepSeek.
The hardware producer, Cerebras, even plans to haunt NVIDIA on directly on the stock market:
>Silicon Valley-based Cerebras, which has filed for an initial public offering that is delayed while U.S. officials review G42's involvement with the firm, is one of the few challengers to Nvidia (NVDA.O), opens new tab for training AI models. But the partnership with Mistral on Thursday is for serving an app based on its model to users, a step called "inference" in the AI industry.
[https://www.reuters.com/technology/artificial-intelligence/ai-chip-firm-cerebras-partners-with-frances-mistral-claims-speed-record-2025-02-07/](https://www.reuters.com/technology/artificial-intelligence/ai-chip-firm-cerebras-partners-with-frances-mistral-claims-speed-record-2025-02-07/)
More information from Cerebras: [https://cerebras.ai/blog/mistral-le-chat](https://cerebras.ai/blog/mistral-le-chat)
You can try out Le Chat with Cerebras here, and see for yourself: [https://chat.mistral.ai/chat](https://chat.mistral.ai/chat) (spoiler: it is fast)
Bad news for the NVIDIA stock in the short- to medium-term, at least.