That is a genuinely big deal for open AI. It is also a 1.56 TB download, and the honest answer to “can I run this on my rig?” is no, not yet, and probably not ever on a single home machine. Both of those things are true at once, so let’s go through what actually shipped, who it is really for, and the benchmark argument that showed up the day before the weights did.

What Moonshot actually released

Per the official model card, K3.1 Pro is a mixture-of-experts model with 2.8 trillion total parameters, of which 104 billion are active for any given token. It has 896 experts, routes each token to 16 of them plus two shared experts, and stacks 93 layers. It takes text and images as input through a 401-million-parameter MoonViT-V2 vision encoder, and its context window is 1,048,576 tokens, a full million.

The architecture is not a brand-new design. Moonshot says K3.1 Pro is post-trained from Kimi K3, with one real structural change: the full-attention layers are now Gated DeepSeek Sparse Attention layers, meant to make long contexts cheaper. That leaves 69 Kimi Delta Attention layers and 24 sparse-attention layers. The rest of the improvement, according to Moonshot’s tech blog, comes from much larger reinforcement learning runs across “hundreds of downstream domains.”

The format detail matters most for anyone thinking about hardware. The weights ship as MXFP4, with MXFP8 activations, and the model was trained that way using quantization-aware training. In plain terms, this 4-bit file is the precision Moonshot serves in production, not a squeezed-down copy. The Hugging Face file listing comes to 96 safetensors shards and about 1.56 TB. ArtificialWatch notes that’s the same size as the Kimi K3 release, which makes sense given the shared shape.

Licensing is friendlier than a lot of people assumed. The Kimi K3.1-Pro License reads like MIT with two conditions. If you run a “Model as a Service” business, such as selling API access, and you pull in more than $20 million over any 12 months, you need a separate agreement with Moonshot. If your product has more than 100 million monthly users or more than $20 million in monthly revenue, you have to show the “Kimi K3.1-Pro” name in your interface. Internal use is exempt from both. For researchers, startups and companies running it on their own infrastructure, it is effectively open.

On capability, Moonshot is pitching this as an agentic coding and work model. Its headline number is 65.7% on Terminal-Bench 4.0, just behind Claude Opus 5.5 at 66.4% and ahead of GPT-6 Astra at 58.2%, per the company’s own table. It also claims the top score in that table on SWE Marathon v1.1, at 51.7. The blog shows off games built in Unity, Unreal and Godot, plus 3D modeling through scripted Blender and CAD work. As a gamer, I find the Godot demo above charming, but keep in mind it’s a curated vendor showcase, not an independent test.

Moonshot’s table also shows the model trailing in places. On the full Humanity’s Last Exam set without tools, K3.1 Pro scores 51.1 against 59.8 for Opus 5.5, and it is clearly behind on several of the cybersecurity tests. That last gap is deliberate. Moonshot says it found and removed cyber-related data from both mid-training and RL for the open release. On its internal xdev-bench-hard exploit benchmark, the unfiltered checkpoint solved 90.5% of problems, while the shipped model solves 81.0%. Moonshot says it doesn’t want open weights to “extend the absolute frontier of cyber capability.” Whether you see that as responsible or as a convenient story, it’s one of the more detailed explanations of an open-weight safety tradeoff any lab has published.

The local AI reality check

Here’s where the local AI crowd, me included, has to be realistic. The vLLM recipe for K3.1 Pro starts at eight data-center GPUs running tensor parallel: two NVIDIA GB300 trays or one B300 node, with a one-node AMD Instinct MI355X setup also listed. The model card recommends vLLM, SGLang or TokenSpeed. Those are server inference engines, not desktop apps.

The usual home route doesn’t exist yet either. llama.cpp shipped v0.9.0 on November 4, the same day the weights dropped, with new support for Mistral Large 4, Metis-1.6 and others, but nothing for K3.1. As ModelFit and Modem Guides both point out, the config declares a new kimi_k31 architecture. llama.cpp’s Kimi K3 converter can’t handle the sparse-attention layers, so nobody can make a GGUF yet. Repositories already carry the K3.1 Pro name, including an Unsloth mirror of the original files, but none of them is a desktop-ready quant.

Even once someone writes support, the floor stays high. Working from Unsloth’s measured Kimi K3 builds, Modem Guides estimates that a 1-bit K3.1 build would need roughly 610 GB of combined RAM and VRAM. That’s beyond even a maxed-out 512 GB workstation, never mind a gaming PC. Some Hacker News commenters have suggested streaming the weights from fast SSDs on smaller machines, but that’s a slow workaround that nobody has benchmarked, not a real plan.

So who is this for? Universities, inference providers, and companies that want frontier-class agents on hardware they control, with no data leaving the building. For everyone else, open weights still help: anyone can host the model, so third-party providers can compete on price and privacy instead of everyone going through Moonshot’s own API. Early reaction rounded up by LLMs Blog found that several hosts listed it on day one.

Moonshot’s strongest pitch isn’t a benchmark. It’s the long-horizon case study in the image above. In an internal test, the company says, K3.1 Pro built a muscle-driven human walking simulator and worked through 167 missions to take the virtual body from collapsing within seconds to walking for 11.2 seconds. Along the way it found and fixed 36 errors in the simulator’s shoulder-muscle math instead of just endlessly tweaking rewards. If that holds up outside Moonshot’s lab, it’s exactly the kind of patient debugging that makes agents useful.

Cody’s take

The benchmark question deserves a clear answer, and right now there isn’t one. On November 3, a day before the weights went public, two researchers at R40 Research posted a preprint audit of K3.1’s behavior. They report that in about 23% of Terminal-Bench 4.0 attempts, the model’s reasoning cited “instructions” that weren’t in the task but that the hidden tests check. They also found that when they edited a task file to contradict one of those hidden rules, K3.1 ignored the edit and failed in three of four cases, while other models followed it. The authors think this most likely points to training on the benchmark tasks themselves, but they are explicit that their analysis “cannot prove it.” I haven’t found a public response from Moonshot.

That doesn’t make K3.1 Pro a bad model. Plenty of developers say K3 and K3.1 feel strong in day-to-day coding, and open weights are exactly what lets independent researchers poke at claims like these. It does mean I’d treat the near-tie with Opus 5.5 on Terminal-Bench as unproven until more independent tests come in.

The bigger picture is still exciting. A 2.8-trillion-parameter model with a million-token context, under a license almost anyone can use, is a real win for open AI, even if most of us will only ever touch it through a hosted endpoint. If you want something you can actually run at home this week, the smaller models in that same llama.cpp release are the place to start. K3.1 Pro is a sign of where open models are heading, and it’s worth watching closely before you trust the scoreboard.