Kimi K3: The Complete Developer Guide
Quick Answer
Kimi K3, developed by Moonshot AI, is the largest open-weight model with 2.8 trillion parameters, designed for advanced tasks like long-horizon coding and deep reasoning.
Quick Take
It introduces innovative features such as Kimi Delta Attention and Mixture-of-Experts sparsity, enhancing its performance and scalability. The model is now available via Together AI's OpenAI-compatible API.
Key Points
- Kimi K3 is the first open-source model in the 3-trillion-parameter class.
- It features Kimi Delta Attention for efficient long-context processing.
- The model supports a maximum context length of 1 million tokens.
- Mixture-of-Experts framework activates only 2% of experts per token.
- API usage is compatible with OpenAI's standards for easy integration.
DeepSignal Analysis
What happened
Moonshot AI has launched Kimi K3, a model with 2.8 trillion parameters, making it the largest open-weight model available. It features advanced capabilities for long-horizon coding and deep reasoning, and introduces architectural innovations like Kimi Delta Attention and Mixture-of-Experts sparsity. The model is accessible via Together AI's OpenAI-compatible API.
Key evidence
- Kimi K3 is the largest open-weight model released, with 2.8 trillion parameters, and is designed for advanced tasks like long-horizon coding.
- The model incorporates Kimi Delta Attention, which supports a context length of 1 million tokens, enhancing its ability to process longer sequences.
- Moonshot AI has implemented Mixture-of-Experts sparsity, activating only 16 out of 896 experts, which poses unique challenges for routing and optimization.
Why it matters
The introduction of Kimi K3 signifies a notable advancement in the field of AI, particularly in open-source models. Its size and capabilities could influence the development of future AI systems, especially in complex reasoning tasks. The architectural innovations may also set new standards for efficiency and performance in large-scale models, potentially impacting various applications in AI development.
Source Excerpt
Kimi K3 is the first open 3T-class model. See how it benchmarks, what it costs, and how to call it on the Together AI API, with copy-paste code examples.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from Together AI
See more →
Open, convenient and predictable: Introducing Provisioned Throughput
Together AI introduces Provisioned Throughput, offering guaranteed inference capacity for MiniMax M3 and GLM-5.2 at $0.05 per PTU per minute, achieving costs up to 90% lower than Claude Opus 4.8. This new model provides predictable pricing and a 99% uptime SLA, catering to companies transitioning to open weight models for production workloads.
