Moonshot, a Chinese AI startup, has released Kimi K3, the largest open-source model at 2.8 trillion parameters, surpassing Anthropic's Fable 5 on a key front-end coding benchmark while offering substantially lower usage costs.
- Kimi K3 tops Fable 5 on front-end coding Arena benchmark
- Model is largest open-source AI at 2.8 trillion parameters
- Pricing is significantly cheaper than proprietary competitors
What happened
Moonshot announced the release of Kimi K3, an open-source AI model with 2.8 trillion parameters, making it the largest of its kind currently available. The model is designed for long-horizon coding, knowledge work, and complex reasoning tasks, showing particular strength in front-end coding as measured by the Arena benchmark.
In head-to-head comparisons, Kimi K3 outperformed Anthropic's Fable 5 in these coding tasks, although its overall performance across all benchmarks is roughly on par with Fable 5 and OpenAI's GPT-5.6. Moonshot plans to publicly release the model's weights by July 27, 2026.
Why it matters
Kimi K3's performance and open-source availability challenge the domain of costly proprietary models dominated by US companies like Anthropic and OpenAI. While Fable 5 charges $50 per million output tokens, Kimi K3's usage is estimated at just $15, highlighting potential industry cost savings.
However, open-source models lack the extensive safety and governance frameworks implemented by proprietary providers, raising concerns about their responsible use. This is compounded by the geopolitical context, as Moonshot is a Chinese firm operating amid ongoing US governmental scrutiny of Chinese technology.
What to watch next
Observers will be closely watching the public release of Kimi K3’s weights and how developers and organizations adopt an open-source model capable of competing with leading proprietary offerings. The trade-offs between cost, performance, and safety frameworks will be critical in influencing user preference.
Additionally, the broader AI ecosystem may see increased debate over international competition and cooperation, especially as US government agencies increasingly engage with AI firms on voluntary safety standards and oversight. The response to Moonshot’s aggressive pricing and open model may also affect future licensing and regulatory policies.