Kimi K3: The Complete Developer Guide

What you'll learnWhat is Kimi K3, and what makes it different?What is under the hood: KDA, Attention Residuals, and the Stable LatentMoE architectureHow do you use reasoning effort, streaming, tools, vision, and 1M context?How do you take it from a first API call to production?How does Kimi K3 compare to the frontier on coding and agentic benchmarks?How much does Kimi K3 cost on Together AI?Kimi K3 is Moonshot AI's most capable model to date: a 2.8-trillion-parameter model and the world's first open-source model in the 3-trillion-parameter class. It is designed for frontier intelligence work like long-horizon coding, end-to-end knowledge work, and deep reasoning. It is also the first open-weights model competing at the GPT 5.6 Sol and Claude Fable 5 tier, and Together AI is working directly with the Moonshot team to serve it.The largest open-weight model releasedThe Kimi team is deeply committed to scaling, and it shows: in nine of the twelve months from July 2025 to July 2026, Kimi models set the upper bound of open-model scale. At 2.8 trillion parameters, K3 is now the largest open-weight model ever released. Available now Run Kimi K3 on Together AI Full 1M context, automatic prefix caching, OpenAI-compatible API, served from US infrastructure. Open the playground What is under the hoodTwo architectural updates form K3's backbone, both designed to help information flow more easily through longer sequences and deeper into the network:Kimi Delta Attention (KDA): a hybrid linear attention mechanism that provides an efficient foundation for scaling attention across very long contexts. This is the first Kimi model to support a 1M context length.Attention Residuals (AttnRes): selectively retrieves representations across model depth rather than accumulating them uniformly.Source: Kimi K3On top of that, Moonshot pushed Mixture-of-Experts sparsity further with the Stable LatentMoE framework, efficiently activating 16 of 896 experts. At this level of sparsity, roughly 2% of experts activated per token, routing and optimization become first-order challenges, so several supporting techniques enable stable training at 2.8T scale:Quantile Balancing: derives expert allocation directly from router-score quantiles, eliminating heuristic updates and a sensitive balancing hyperparameter.Per-Head Muon: extends the Muon optimizer to optimize attention heads independently for more adaptive learning at scale.Sigmoid Tanh Unit (SiTU): improves activation control.Gated MLA: improves attention selectivity.How to use Kimi K3 on Together AIThe API is OpenAI-compatible. The snippets below target Together AI and use the official Together Python SDK. python3 -m pip install --upgrade 'together>=2.0.0' import os from together import Together MODEL = "moonshotai/Kimi-K3" client = Together( api_key=os.environ["TOGETHER_API_KEY"], ) completion = client.chat.completions.create( model=MODEL, messages=[{"role": "user", "content": "Introduce Kimi K3 in one sentence."}], max_tokens=130_000, ) print(completion.choices[0].message.content) Thinking effortK3 can be configured with the top-level reasoning_effort field. Three levels are supported: low, high, and max, with max as the default. On Together, thinking can also be switched off via the standard reasoning={"enabled": False} toggle. # Adjust depth: "low" | "high" | "max" completion = client.chat.completions.create( model=MODEL, reasoning_effort="max", messages=[{"role": "user", "content": "Prove that the square root of 2 is irrational."}], max_tokens=8192, ) # Instant mode, no thinking tokens billed at all fast = client.chat.completions.create( model=MODEL, messages=[{"role": "user", "content": "What is the capital of France?"}], reasoning={"enabled": False}, max_tokens=256, ) StreamingStreaming responses deliver separate reasoning_content (t