Moonshot AI just released Kimi K3, a 2.8 trillion-parameter open-weight model that immediately slots into conversations previously reserved for the biggest closed systems from US labs. The Beijing-based startup launched the model on July 16 at the World Artificial Intelligence Conference in Shanghai, where Chinese President Xi Jinping used the occasion to call for international cooperation on AI development.
Kimi K3 is the first open model to enter what researchers are calling the 3-trillion parameter class, and early benchmarks suggest it can hang with the likes of Claude Fable 5 and GPT-5.6 Sol.
What makes Kimi K3 different
The model uses a Mixture-of-Experts (MoE) architecture. Instead of running all 2.8 trillion parameters on every query, the system activates only a subset of its 896 experts depending on the task. It also features something Moonshot is calling “Kimi Delta Attention” and “Attention Residuals,” proprietary techniques that appear to improve how the model handles long sequences of information.
Kimi K3 supports a 1 million-token context window. Moonshot describes its target use cases as “long-horizon coding” and knowledge work productivity. The model also ships with native multimodal capabilities, including 3D generation. On LMArena’s Frontend Code Arena benchmark, Kimi K3 debuted at the number one spot, with a third-place overall ranking across broader evaluations.
Moonshot made the API available immediately at launch. Full model weights are scheduled for release on July 27, giving developers and researchers roughly ten days to access the model via API before they can run it locally.
The geopolitical subtext
Xi Jinping’s appearance at the Shanghai conference was not coincidental. The Chinese president used his remarks to emphasize open-source AI development as a vehicle for international cooperation, a framing that positions China as the generous counterpart to US firms that keep their most powerful models locked behind proprietary walls.
US AI policy has been characterized by export controls on advanced chips and growing restrictions on technology transfers to China. This follows the pattern set by DeepSeek earlier, which demonstrated that Chinese AI labs could produce competitive models at a fraction of the cost assumed by Western incumbents.
What this means for crypto and tech investors
There are no tokens associated with Kimi K3. No governance token, no utility token, no hastily assembled DAO. But the release of a model this powerful under an open-weight license puts immediate pressure on every AI project that has built its value proposition around model access or proprietary intelligence.
Decentralized GPU networks like Render, Akash, and io.net could see increased demand as developers look to run a 2.8 trillion-parameter model without paying hyperscaler prices.
Watch how the benchmark conversation evolves once full model weights drop on July 27. That’s when independent researchers will be able to verify Moonshot’s claims and stress-test the model on real-world tasks.