Moonshot’s Kimi K3.1 Pro Takes Aim at AI’s Top Models

Moonshot has introduced Kimi K3.1 Pro, a major upgrade to its Kimi K3 model released back in July. The new system targets the work where advanced AI models face their hardest tests: complex coding, financial research, 3D creation, software security, and tasks that require repeated revisions.

The headline number is striking. Kimi K3.1 Pro is a 3-trillion parameter post-trained version of Kimi K3, and third-party testing places it near the top of the current model race. Moonshot says it is “matching the strongest proprietary models,” while Artificial Analysis placed it No. 4 overall on its Intelligence Index.

A strong showing against closed models

Artificial Analysis ranks Kimi K3.1 Pro ahead of Xiaomi’s MiMo-V2.6-Pro, DeepSeek V4.1-Flash-1017, Meta’s new Muse Wonder, and OpenAI’s GPT-6 Astra. The model trails only Anthropic’s 5.5 duo and Meta’s Muse Wonder on that ranking, which puts it ahead of GPT-6 Astra despite being positioned as an openly available model.

Moonshot’s own testing paints an even stronger picture. The company says Kimi K3.1 Pro tops Opus 5.5 on hours-long coding projects, expert academic questions, document-heavy research, and real-world job tasks. Its largest gains show up in work that involves repeated coding, tool use, debugging, and revision rather than a single answer.

That distinction matters because many useful AI tasks do not end after the first response. A model may need to inspect a problem, write code, test it, find an error, revise the result, and repeat the process. Kimi K3.1 Pro is designed to handle that cycle, along with financial analysis, 3D modeling, and software-security work.

Artificial Analysis also ranks Kimi K3.1 Pro as the best openly available cyber-defense model. It scored 64% with no safety refusals, trailing only a GPT-6 variant available to vetted defenders. The result takes on added weight after Dario Amodei gave defenders 6-12 months in May before Chinese models caught up. K3.1 Pro is already near the top on defense.

Finance, 3D creation, and game development

Moonshot reports a 7.2-point gain in financial modeling and an 11.6-point gain in research. The company also says the model scored approximately 4.2 times its predecessor’s result in a simulated business-management exercise.

Those improvements cover spreadsheet modeling, investment screening, profitability analysis, and research that traces conclusions back to evidence. That combination gives Kimi K3.1 Pro a clear focus on tasks where users need both calculations and explanations for how the answer was reached.

The model’s 3D abilities extend beyond simple image generation. Moonshot says Kimi can turn a written brief, sketch, or reference image into a model, work with existing asset collections, and modify geometry, materials, and scene arrangements through Blender or CAD scripts. Those scripted operations do not require MCP connectors.

Moonshot highlights adjustable designs for mechanical components, architecture, configurable products, and furniture. Its coding showcase explicitly includes Unreal Engine alongside Unity and Godot, and the company says Kimi can combine game-engine work with web research and 3D generation to produce complete interactive experiences.

The 3D Make application lets users inspect and edit models online, download them for 3D printing, or export videos and Live Photos. That creates a path from an idea to a usable model without keeping every step inside a text conversation.

Pricing, access, and the open-weights plan

Developers and users can access Kimi K3.1 Pro through Moonshot’s applications and developer API. The model costs $2 in and $10 out per million tokens, a price Moonshot positions like Sonnet 5.5. Its rates are half those of Claude Opus 5.5 and Meta’s Muse Wonder for input and output, while DeepSeek and Xiaomi remain cheaper.

Another claim puts the price at a fifth of Astra’s rate. The trade-off is token use: Kimi K3.1 Pro consumes tokens at a higher rate than its frontier rivals, so the headline price may not tell the whole story for long, complex tasks.

Moonshot says it will release the weights on November 4, 2026, under an open license. That release would make the model free to download and could strengthen its position as the top open-weights AI model in the world, especially after its showing on Artificial Analysis benchmarks.

The launch has not been free of controversy. Kimi K3.1 Pro has been the subject of an alleged paid influencer scandal on X, the social network. That allegation sits alongside the model’s benchmark results and does not change the basic access plan: users can try it through Moonshot’s applications and API now, with the weights planned for November 4, 2026.

What this means for everyday AI work

The broader appeal of systems like Kimi K3.1 Pro is their ability to connect research, creation, and action. The Rundown’s weekly Roundtable describes how Jonah, its Head of AI Education, and his wife used ChatGPT to bring a new master bedroom vision to life instead of hiring a designer.

They began by sketching the basic layout, which ChatGPT rendered into a floor plan. They identified the wood, the type of cut, and the finishes, then held a detailed conversation about how the room should look from each side. For smaller details, they scanned Google Images and Pinterest for inspiration pictures and designs.

They used ChatGPT Voice for back-and-forth conversations and final tweaks, re-rendering the images as their ideas developed. The finished images went to their contractor to explain the higher-level vision they wanted to achieve.

Jennifer, a Tech & Robotics Writer, described another research use. She is narrowing down high school options for her daughter, who wants to focus on art, and uses Astra to find public schools with strong art programs in her area and beyond.

Astra can compare school rankings and academic results, review art courses and facilities, and bring admissions requirements, portfolio expectations, fees, and commute times into one comparison. It can also suggest preparation classes, including drawing, painting, digital art, and portfolio workshops, then help organize open days, application deadlines, and questions for school visits.

These examples show why coding, research, design, and secure tool use matter in the model race. The best system is not only the one that answers a question. It is the one that can help users develop an idea, check evidence, revise work, and finish a project.

Stay connected via Google News

Follow us for the latest AI News