Kimi K3: Moonshot AI Unveils a 2.8-Trillion-Parameter Open-Weight Model

Introduction

Moonshot AI has introduced Kimi K3, a new multimodal reasoning model with 2.8 trillion total parameters and a 1-million-token context window.

The Beijing-based company describes Kimi K3 as the first open model to reach the 3-trillion-parameter class. It is designed for long-running software engineering, knowledge work, visual reasoning, research, and multi-step agent tasks.

The scale immediately attracted attention. Kimi K3 is considerably larger by total parameter count than the trillion-parameter open-weight systems that defined the previous generation of Chinese frontier models.

Size, however, is only one part of the story. Kimi K3 uses a highly sparse Mixture-of-Experts architecture, activating only a small portion of its experts for each token. Moonshot also introduced architectural changes intended to improve how information moves across long sequences and deep networks.

The company says Kimi K3 reaches frontier-level performance across many of its tests, although it still trails the strongest proprietary models overall. Independent evaluations released around the launch also place it among the leading currently available models.

A 2.8-Trillion-Parameter Multimodal Model

Kimi K3 is a natively multimodal model capable of working with text and visual input. Moonshot positions it for tasks that require sustained reasoning rather than short, isolated responses.

Its main specifications are:

Specification	Kimi K3
Total parameters	2.8 trillion
Architecture	Sparse Mixture of Experts
Routed experts	896
Experts activated per token	16
Context window	1 million tokens
Modalities	Text and visual input
Main use cases	Coding, knowledge work, research, reasoning, multimodal creation
Default reasoning mode at launch	Maximum thinking effort
Full-weight release target	By July 27, 2026
Recommended large-scale deployment	Supernode configurations with at least 64 accelerators

The total parameter count describes the full model, not the number of parameters used for every token. Kimi K3 activates 16 of its 896 experts during inference, allowing it to benefit from a very large capacity without running the entire network for each generated token.

This distinction is important. A sparse model can contain trillions of parameters while keeping active computation substantially below what a dense model of the same total size would require.

Even with sparse activation, Kimi K3 remains an exceptionally demanding system to host. Moonshot recommends deployment on supernode configurations with 64 or more accelerators, placing practical self-hosting beyond the reach of most individual developers and smaller teams.

The Architecture Behind Kimi K3

Moonshot highlights two core architectural technologies: Kimi Delta Attention and Attention Residuals.

Kimi Delta Attention

Kimi Delta Attention, or KDA, is designed to make attention more efficient as context length increases.

Traditional attention becomes expensive when a model processes very long sequences. KDA is intended to provide a more scalable foundation for handling large contexts while retaining the information needed for reasoning and tool use.

The 1-million-token window gives Kimi K3 enough capacity to work with large repositories, extensive document collections, long agent histories, or combinations of source material that would exceed the limits of many earlier models.

A large context window does not guarantee that every token will be used equally well. Retrieval quality, reasoning stability, prompt structure, and the agent harness remain important. Still, the expanded window gives developers more room to keep relevant material available without aggressive early compression.

Attention Residuals

Attention Residuals, abbreviated as AttnRes, changes how representations are carried through the depth of the model.

Instead of allowing information to accumulate uniformly through conventional residual connections, the mechanism selectively retrieves useful representations from earlier layers.

Moonshot says this improves information flow across model depth and contributes to more efficient scaling.

Stable LatentMoE

Kimi K3 combines sparse expert routing with a Stable LatentMoE framework. Only 16 of 896 routed experts are activated for a token.

At this level of sparsity, routing quality becomes a major engineering problem. A poorly balanced system can overload a subset of experts, waste hardware, or reduce throughput.

Moonshot says it uses quantile-based balancing and a balanced expert-parallel training method to keep workloads distributed efficiently without host synchronization in the critical path.

Quantization-Aware Training

The model uses quantization-aware training from the supervised fine-tuning stage onward.

Moonshot reports that Kimi K3 uses MXFP4 weights and MXFP8 activations. Lower-precision formats can reduce memory and communication requirements, although production performance still depends on hardware and inference-engine support.

Together, these architectural changes are said to provide roughly 2.5 times better overall scaling efficiency than Kimi K2.

Designed for Long-Horizon Coding

Coding is one of the main areas Moonshot emphasizes for Kimi K3.

The model is intended to remain productive across extended engineering sessions, move through large repositories, operate terminal tools, inspect runtime feedback, and continue iterating with limited human supervision.

This differs from a model that only generates a single function from a prompt. Long-horizon coding may require the system to:

1. Explore an unfamiliar repository.
2. Understand architecture and dependencies.
3. Form a multi-stage plan.
4. Edit files across several modules.
5. Run builds and tests.
6. Read error messages and logs.
7. Compare screenshots or rendered output.
8. Revise the implementation repeatedly.
9. Preserve relevant context throughout the task.

Moonshot also emphasizes tasks that combine coding with visual reasoning. Kimi K3 can use screenshots and visual feedback while working on frontend interfaces, games, CAD workflows, and other interactive software.

Moonshot’s Coding Benchmark Results

Moonshot evaluated Kimi K3 across several software-engineering benchmarks and agent harnesses.

The company reports strong results in:

• DeepSWE.
• Terminal-Bench 2.1.
• FrontierSWE.
• Program Bench.
• SWE Marathon.
• Kimi Code Bench 2.0.
• GPU kernel optimization.
• Compiler construction.

These evaluations do not all use identical setups. Some models were tested through Kimi Code, others through Claude Code or Codex. Moonshot documents these differences in the benchmark footnotes.

That matters because the agent harness can influence results. Tool interfaces, context management, retry behavior, permission systems, and prompt design may all affect how well a model performs on a long-running task.

Moonshot reports that Kimi K3:

• Scored 67.5 on its DeepSWE comparison.
• Reached 86.3 on Terminal-Bench 2.1.
• Scored 81.2 on FrontierSWE.
• Reached 77.8 on Program Bench.
• Scored 42.0 on SWE Marathon.
• Reached 72.8 on the internal Kimi Code Bench 2.0.

These are company-reported results and should be read alongside independently operated benchmarks.

GPU Kernel Optimization

Moonshot also tested whether the model could optimize GPU kernels rather than only write conventional application code.

Each evaluated model was placed in an equivalent sandbox and given as long as 24 hours to profile, rewrite, and benchmark several tasks. The workloads included components related to Kimi K3’s own architecture and were tested across NVIDIA H200 hardware and another general-purpose GPU platform.

Moonshot says Kimi K3 performed competitively with Claude Fable 5 when fallback behavior was allowed and substantially outperformed several other tested models in this setup.

The company also says an early Kimi K3 checkpoint handled much of the kernel-optimization work used during the model’s own development.

This capability matters because efficient AI systems depend on more than model architecture. Kernel engineering determines how effectively the model uses accelerators, memory bandwidth, communication links, and specialized numerical formats.

Building a GPU Compiler from Scratch

In another internal test, Kimi K3 created a compact Triton-like GPU programming system called MiniTriton.

According to Moonshot, the resulting project included:

• A tile-level intermediate representation built on MLIR.
• Optimization passes.
• A PTX code-generation pipeline.
• A domain-specific language frontend.
• Runtime components.
• Support for end-to-end nanoGPT training.

Moonshot reports that MiniTriton matched or exceeded Triton and torch.compile on selected supported benchmarks while maintaining stable training behavior in its nanoGPT test.

This result should be interpreted as a company case study rather than proof that a generated compiler is ready to replace mature production infrastructure. It does, however, illustrate the type of multi-stage engineering task Moonshot designed Kimi K3 to handle.

Visual Coding, Game Development, and Chip Design

Kimi K3’s native multimodal design allows it to alternate between source code and visual output.

For game and frontend work, the model can generate an implementation, inspect a screenshot, identify visual problems, and revise the code. Moonshot refers to this as keeping vision inside the development loop.

The company also presented an early chip-design demonstration. In a 48-hour autonomous run, Kimi K3 reportedly designed, optimized, and verified a chip for serving a small model based on its own architecture.

The proof of concept used open-source electronic design automation tools and the Nangate 45 nm library. Moonshot reports that the simulated design:

• Fit within 4 square millimeters.
• Closed timing at 100 MHz.
• Included about 1.46 million standard cells.
• Used 0.277 MB of SRAM.
• Contained an INT4 multiply-accumulate array with fused dequantization.
• Reached more than 8,700 tokens per second in simulation.

This was a simulation and early research demonstration, not a manufactured production chip.

Kimi K3 for Scientific and Knowledge Work

Moonshot is also positioning Kimi K3 as a model for research and professional knowledge workflows.

In one computational astrophysics example, the model was asked to reproduce universal relations involving neutron-star properties.

Moonshot says Kimi K3:

• Reviewed and cross-checked more than 20 papers.
• Implemented a complete numerical workflow.
• Evaluated over 300 equations of state.
• Identified inconsistencies in published formulas.
• Generated more than 3,000 lines of Python.
• Built an interactive HTML dashboard.

The company estimates that the project took the model about two hours, compared with one or two weeks for an experienced researcher.

As with the other case studies, the result should not be treated as a universal time-saving guarantee. Performance depends on the task, available tools, source quality, validation method, and the level of human review.

A 1-Million-Token Context Window

Kimi K3’s context window is one of its most commercially relevant specifications.

One million tokens can accommodate combinations such as:

• A large software repository with documentation and logs.
• Many lengthy reports and source documents.
• Extended tool-call histories.
• Long-running research sessions.
• Large collections of images and text.
• Multiple drafts, revisions, and intermediate results.

Moonshot says Kimi K3 performs best when its complete reasoning history is preserved.

This creates an important implementation requirement. If an agent harness drops parts of the model’s earlier thinking history, summarizes it incorrectly, or switches from another model midway through a session, output quality may become unstable.

Moonshot recommends using a verified compatible harness such as Kimi Code and avoiding model switching during an active K3 session.

Independent Benchmark Results

Independent tests broadly support the conclusion that Kimi K3 is a frontier-class model, although they do not show it leading every category.

Artificial Analysis gave Kimi K3 a score of 57 on its Intelligence Index and placed it third at the time of evaluation. Its overall result was described as comparable with Claude Opus 4.8 and GPT-5.5, while remaining behind Claude Fable 5 and GPT-5.6 Sol.

Artificial Analysis also reported that Kimi K3 was unusually verbose during its evaluation, generating around 130 million tokens compared with an average of approximately 63 million for comparable models.

That matters for cost and latency. A model can be competitively priced per token while still becoming expensive if it uses substantially more tokens to complete the same task.

Reuters reported additional third-party results around launch:

• Arena.ai ranked Kimi K3 first in a web-interface-building evaluation.
• Vals AI placed it second overall behind Claude Fable 5 and ahead of GPT-5.6 Sol.
• Artificial Analysis found performance comparable to GPT-5.5 and Claude Opus 4.8 on complex, multi-step work.

Benchmark rankings change as evaluators add samples, update harnesses, and retest models. Current leaderboards should therefore be checked before making a purchasing or deployment decision.

Open-Weight Status and Release Timing

Kimi K3 was announced as an open-weight model, but the full downloadable weights were not yet available when this article was prepared.

Moonshot’s official release page states that the weights are scheduled to be published by July 27, 2026.

This distinction is important:

• The model, architecture, API, and product integrations have already been announced.
• Kimi K3 is already accessible through Moonshot’s hosted products and API.
• The full model weights are planned for a later date.
• Local deployment instructions and final ecosystem support may continue to evolve after the weight release.

Until the weights are published with their license and technical files, claims about exact self-hosting procedures, supported inference engines, or hardware configurations should be treated as preliminary.

Open-weight also does not automatically mean unrestricted open source. The eventual license will determine what users may modify, redistribute, or use commercially.

Why Most Users Will Not Run Kimi K3 Locally

A 2.8-trillion-parameter model is far beyond the normal definition of local AI.

Even with sparse activation and low-precision weights, the full checkpoint, expert routing, high-bandwidth communication, context cache, and runtime overhead require a large distributed system.

Moonshot recommends at least 64 accelerators in a supernode-style configuration for deployment.

The practical implications are:

1. Individual users are more likely to access Kimi K3 through a hosted API or application.
2. Enterprises may use specialized inference providers rather than maintain the cluster themselves.
3. Open weights are most immediately useful to well-funded research laboratories, cloud platforms, inference companies, and large infrastructure teams.
4. Quantized or distilled community variants may eventually lower the barrier, but those versions may not match the full model.

Reuters cited an estimate that self-hosting the system could require computing equipment worth hundreds of thousands of dollars.

The exact cost will depend on the final weight format, accelerator type, networking, context length, throughput target, redundancy, and serving software.

API Pricing and Availability

Kimi K3 is already available through several Moonshot products:

• Kimi.com.
• Kimi Work.
• Kimi Code.
• Kimi API.
• Kimi Enterprise.

Kimi Code users can select Kimi K3 through the model command in the terminal interface.

Moonshot’s current API pricing is:

Token category	Price per 1 million tokens
Cache-hit input	$0.30
Cache-miss input	$3.00
Output	$15.00

The API model name is:

kimi-k3

Moonshot says its official API achieves a cache-hit rate above 90% in coding workloads through the Mooncake disaggregated inference architecture.

Actual application cost will depend on prompt size, reasoning length, cache reuse, tool calls, retries, and output verbosity.

Faster Release Cycles Across Chinese AI Labs

Kimi K3 arrives during a period of unusually fast model releases from Chinese AI companies.

Moonshot, Z.ai, MiniMax, DeepSeek, Meituan, and other developers have been releasing larger and more capable systems while competing heavily on price, open weights, coding, and agent performance.

Reuters noted that this pace is challenging earlier assumptions that leading Chinese models were consistently several months behind U.S. frontier systems.

Model size alone does not prove leadership. Parameter counts are difficult to compare across dense and sparse architectures, and several major U.S. developers do not disclose the size of their current models.

Still, Kimi K3 shows that Chinese laboratories are operating at the frontier of model scale, architecture, long-context inference, and agent engineering.

Market Reaction and Moonshot’s Expansion

Reuters reported that shares of listed Chinese AI competitors fell after the Kimi K3 announcement.

Near the close of trading, Zhipu was down 27.7%, while MiniMax had fallen 16.5%.

A single trading session does not establish the long-term competitive impact of a model release. The reaction does show how closely public markets are watching technical progress among Chinese AI companies.

Moonshot is backed by major investors including Alibaba and Tencent.

Reuters also cited a Bloomberg report saying that Moonshot had been seeking approximately $2 billion in new funding at a valuation of around $30 billion, ahead of a possible Hong Kong listing.

Those financing and listing plans remain reported rather than finalized.

Kimi K3’s Published Limitations

Moonshot’s official release is unusually direct about several limitations.

Sensitivity to Reasoning History

Kimi K3 was trained with preserved reasoning history. If an agent framework fails to return the full historical thinking content, performance may become unstable.

Switching to Kimi K3 midway through a session started with another model is also discouraged.

Excessive Proactiveness

The model was optimized for demanding, long-running tasks.

As a result, it may make unexpected decisions when instructions are ambiguous or when it encounters a minor obstacle. Applications that require narrow behavior should provide explicit constraints through the system prompt or an AGENTS.md file.

User-Experience Gap

Moonshot says that Kimi K3 remains behind Claude Fable 5 and GPT-5.6 Sol in overall user experience, despite being highly competitive on many technical evaluations.

High Infrastructure Requirements

Open weights do not make the full system easy to deploy. Its scale, networking requirements, and long-context memory demands create a high operational barrier.

High Token Usage

Independent testing suggests that Kimi K3 may use considerably more tokens than the average frontier model on complex evaluations. This can affect total cost, speed, and rate-limit consumption.

What the Release Means for Developers

For most developers, Kimi K3 will initially be a hosted model rather than a self-hosted one.

Its most relevant features are:

• A very large context window.
• Strong long-running coding behavior.
• Native visual reasoning.
• Agent-oriented tool use.
• Competitive performance on professional knowledge tasks.
• API access at a published price.
• Planned open-weight availability.

Teams evaluating it should test real workflows rather than rely only on headline benchmark numbers.

Useful evaluations may include:

1. Repository navigation and issue resolution.
2. Long-context instruction retention.
3. Frontend implementation from visual references.
4. Tool-call reliability.
5. Output verbosity and cost.
6. Behavior under ambiguous instructions.
7. Compatibility with the intended agent harness.
8. Performance when reasoning history is compacted.
9. Data-governance and deployment requirements.
10. Recovery from failed commands or tests.

Frequently Asked Questions

What is Kimi K3?

Kimi K3 is Moonshot AI’s 2.8-trillion-parameter multimodal reasoning model. It is designed for long-horizon coding, knowledge work, scientific research, visual tasks, and agent workflows.

Is Kimi K3 open source?

Moonshot describes Kimi K3 as an open model and plans to release its full weights. Open-weight availability is not necessarily the same as open-source software, and the final license will determine the permitted uses, modifications, and redistribution terms.

Are the Kimi K3 weights available now?

Not yet at the time this article was prepared. Moonshot’s official announcement says the full weights will be released by July 27, 2026.

How many parameters does Kimi K3 use during inference?

The complete model contains 2.8 trillion parameters, but it uses a sparse Mixture-of-Experts design. Moonshot says 16 of 896 experts are activated for each token, so the whole network is not active at once.

Can Kimi K3 run on a personal computer?

The full model is not realistically designed for a normal personal computer. Moonshot recommends supernode deployments with at least 64 accelerators, so most users will access Kimi K3 through hosted products or an API.

What is Kimi K3’s context-window size?

Kimi K3 supports a context window of 1 million tokens. The model is sensitive to how its reasoning history is preserved, so agent-framework compatibility and context management are important.

How much does the Kimi K3 API cost?

Moonshot lists cache-hit input at $0.30 per million tokens, cache-miss input at $3.00 per million tokens, and output at $15.00 per million tokens. Total cost also depends on output length, cache reuse, tool calls, and retries.

Is Kimi K3 better than GPT-5.6 Sol or Claude Fable 5?

The answer depends on the task and benchmark. Moonshot says Kimi K3 trails those two models overall while performing competitively or leading on selected coding and knowledge-work evaluations. Independent testing also places Kimi K3 among the leading models but not consistently first.

Related Tools

• Kimi: Moonshot AI’s hosted agent workspace for using Kimi K3.
• Kimi Work: A desktop environment for Kimi’s document, research, dashboard, and knowledge-work features.
• Kimi Code: Moonshot AI’s terminal and IDE coding agent with Kimi K3 model selection.
• Kimi API Platform: Official API access, pricing, developer documentation, and account management.
• Moonshot AI on GitHub: Official open-source projects, model tooling, and research repositories.
• Moonshot AI on Hugging Face: Official model cards and downloadable Moonshot AI model releases.
• vLLM: An open-source inference engine that Moonshot says will receive KDA prefix-cache support.

Related Links

• Kimi K3 Official Technical Blog: Moonshot AI’s architecture, benchmark, availability, and limitations announcement.
• Moonshot AI Official Website: The company’s official product and research homepage.
• Kimi API Platform: Official Kimi K3 API access and current token pricing.
• Moonshot AI GitHub Organization: Official repositories for Kimi models, Kimi Code, and infrastructure research.
• Moonshot AI Hugging Face Organization: Official model-weight and model-card releases.
• Artificial Analysis: Kimi K3: Independent intelligence, pricing, speed, and context-window evaluation.
• Artificial Analysis Kimi K3 Benchmark Report: An independent summary of Kimi K3’s Intelligence Index performance.

Summary

Kimi K3 moves Moonshot AI into the 3-trillion-parameter class with a sparse multimodal model built for coding, research, knowledge work, and long-running agent tasks. Its 1-million-token context window, Kimi Delta Attention, Attention Residuals, and highly sparse expert routing are intended to make that scale more efficient.

Company benchmarks and independent evaluations place Kimi K3 among the current frontier models, although Moonshot acknowledges that it still trails Claude Fable 5 and GPT-5.6 Sol overall. Its high token usage, sensitivity to reasoning history, proactive behavior, and substantial infrastructure requirements are important practical limitations.

Kimi K3 is already available through Moonshot’s applications and API. The complete weights were scheduled for release by July 27, 2026, so most developers will initially use the hosted version rather than deploy the full model themselves.

The key development is not parameter count alone: Kimi K3 combines extreme scale, sparse activation, million-token context, multimodal input, and long-horizon agent behavior in a model that Moonshot intends to release with downloadable weights.