Chinese AI firm Moonshot AI has launched the Kimi K3.1 Pro, a 3-trillion-parameter model featuring a one-million-token context window. Alongside the high-speed version, this model aims to enhance various complex AI tasks and is set for full public weights release by early November.
- 3-trillion-parameter model with 1M-token context window
- Performance surpasses GPT-6 Astra on coding benchmarks
- Full model weights to be publicly available by November 4
What happened
Moonshot AI has introduced its latest AI advancement, the Kimi K3.1 Pro, a model boasting 3 trillion parameters and an extensive one-million-token context window. Along with this full-featured model, a high-speed version offering 2 to 3 times faster interactions was also launched. These models are currently accessible on multiple Moonshot platforms including Kimi.ai, Kimi Work, Kimi Code, and through the Kimi API.
The K3.1 Pro marks a significant upgrade over its predecessor, Kimi K3, benefiting from increased computing power during post-training. It supports a broad array of applications such as comprehensive coding tasks—from minor fixes to large-scale project builds—3D content creation, and detailed financial workflow execution.
Why it matters
The launch of Kimi K3.1 Pro highlights China’s growing presence in developing cutting-edge open-weight AI models that rival leading global proprietary systems. Its performance on benchmarks like Terminal-Bench 4.0, where it scored 65.7%, places it ahead of notable models like GPT-6 Astra, showcasing significant advances in coding AI capabilities.
Moreover, Moonshot AI's enhancements to reinforcement learning frameworks, long-context handling, and infrastructure efficiency enable the K3.1 Pro to perform well in complex autonomous tasks and knowledge-intensive operations, broadening the scope of AI applications that can benefit from open-access models.
What to watch next
Moonshot AI plans to publicly release the full model weights of Kimi K3.1 Pro by November 4, potentially accelerating adoption and innovation in open-weight AI communities worldwide. This move will allow developers, researchers, and enterprises to explore the model’s capabilities deeply and integrate it into diverse applications.
Following the release, attention will focus on how the high-speed variant performs in real-world use cases demanding rapid responsiveness. Additionally, monitoring competitive advancements and how Moonshot AI's technologies influence the landscape of AI in sectors such as finance, software development, and autonomous systems will be crucial.