Moonshot’s Kimi 3 aims to rival Anthropic’s Opus 4.8
▼ Summary
– Moonshot AI’s upcoming Kimi K3 model is expected to match or surpass Anthropic’s Opus 4.8 and will be the largest open-weight AI model from China, with 2 to 3 trillion parameters.
– Kimi K2 models have performed well in the open source market, ranking high on benchmarks and nearly matching frontier models.
– Moonshot is raising fresh capital in a round valuing it at $31.5 billion, following a $2 billion raise at a $20 billion valuation in May.
– Industry leaders debate the value of paying for expensive, closed-source AI models from labs like OpenAI and Anthropic, fearing data extraction.
– Some executives promote cheaper open source models, such as those from DeepSeek or Moonshot, as alternatives, especially as Chinese open models close the gap with frontier models.
Chinese AI startup Moonshot AI is preparing to release its next-generation model, Kimi K3, which sources tell the Financial Times could match or even outperform Anthropic’s Opus 4.8. The news signals a major escalation in the race between open-weight and proprietary AI systems.
The company’s existing Kimi K2 series has already carved out a strong reputation in the open source community, consistently ranking high on key benchmarks and proving that its capabilities are competitive with the latest frontier models. Now, Moonshot aims to close the remaining gap with closed-source leaders like OpenAI and Anthropic.
According to the report, Kimi K3 will be the largest open-weight AI model ever produced in China, boasting a parameter count between 2 trillion and 3 trillion. It is expected to launch in the coming days.
Alongside this technical milestone, Moonshot is reportedly raising new capital at a valuation of $31.5 billion. The company previously secured $2 billion in May at a $20 billion valuation.
This development arrives amid growing skepticism about the value proposition of expensive, closed-source AI models from labs like OpenAI and Anthropic. Industry executives are increasingly concerned that these companies could extract and repurpose client data submitted through products like ChatGPT and Claude.
In response, many leaders are now pitching their own models as safer alternatives, or advising companies to adopt cheaper open-source models from Chinese developers like DeepSeek, Z.ai, or Moonshot, then fine-tune them for proprietary use. This argument has gained significant traction as open models from China continue to narrow the performance gap with their more costly frontier counterparts.
(Source: TechCrunch)
