Kimi K3: The Largest Open Model Ever, and Why Almost No One Can Run It Locally

Moonshot AI announced Kimi K3, a 2.8-trillion-parameter open model, on July 16, 2026, claiming top-tier performance but releasing only an API initially, with full weights promised by July 27. Independent tests rank it near leaders like Claude Opus 4.8 and GPT-5.5, but its massive size makes local deployment impractical for most users.

On July 16, 2026, Moonshot AI announced Kimi K3 , and called it the first open model to cross 2.8 trillion parameters . Independent testers put it near the top of the field, and the press ran with "largest open-source model ever." For a local-LLM site the news is genuine and the timing is awkward: as of the announcement the weights are not downloadable. Moonshot says the full weights land by July 27, 2026 , and until then Kimi K3 is an API and a chat window, not a file you can pull. So this is a preview, not an owner report. Nobody outside Moonshot has run the weights, there is no technical report yet, and there are no community quants. We have not tested it. What we can do is lay out what Moonshot claims, what the one independent eval so far shows, and the part that matters most here: what a 2.8-trillion-parameter model would take to run at home. The short version is that the machine most people would reach for, a maxed-out Mac Studio, cannot hold it at any quant. What Moonshot says Kimi K3 is Straight from Moonshot's announcement https://www.kimi.com/blog/kimi-k3?ref=vettedconsumer.com and its API docs https://platform.kimi.ai/docs/guide/kimi-k3-quickstart?ref=vettedconsumer.com , all of it vendor-stated: | Spec | Kimi K3 as stated by Moonshot | |---|---| | Total parameters | ~2.8 trillion Mixture-of-Experts | | Active per token | Not disclosed. Only the routing is given: "16 of 896 experts" active "Stable LatentMoE" | | Context window | 1,000,000 tokens | | Architecture | Kimi Delta Attention a hybrid linear attention plus Attention Residuals | | Modalities | Native vision text + images + video in , always-on "max thinking" | | Precision | Quantization-aware training in MXFP4 weights / MXFP8 activations | | License | Not stated for K3. K2 and K2.5 shipped Modified MIT, so that is the expectation, not a confirmed fact | | Available now | API + web only model id kimi-k3 ; weights "by July 27, 2026" | Two of those lines deserve a flag. The license is genuinely unknown : the announcement names none, and the "Modified MIT" label floating around comes from aggregator sites inferring it from the K2 family, not from Moonshot. And the active-parameter count is not published , which is the single most important number for anyone thinking about running it. More on why below. One detail local runners should note: Moonshot says K3 was trained quantization-aware in MXFP4 , the same 4-bit block format OpenAI's gpt-oss models use. If that holds through the weight release, the "native" quant is 4-bit-class, which sets the floor on how small the download can get without extra damage. The benchmarks: read them as claims Because the weights are not out, every per-benchmark number Moonshot published is a vendor self-report that no one can reproduce yet. Moonshot's own table leads with agentic and coding results: it claims 91.2 on BrowseComp which it frames as state-of-the-art at release , 93.5 on GPQA-Diamond , and 88.3 on Terminal-Bench 2.1 . Take those as the maker's numbers, not settled facts. Note also there is no SWE-bench Verified score for K3 from any credible source, so any figure you see claiming one is invented. The number worth trusting is the independent one. Artificial Analysis https://artificialanalysis.ai/models/kimi-k3?ref=vettedconsumer.com , which runs its own evals through the API, scored Kimi K3 at 57 on its Intelligence Index and ranked it third to fourth overall: on par with Claude Opus 4.8 and GPT-5.5, a step behind the current leaders Fable 5 and GPT-5.6 Sol . That directionally backs Moonshot's positioning. Separately, blind human-preference testing on frontend coding relayed by Axios https://www.axios.com/2026/07/16/moonshot-kimi-ai-china-model-openai-anthropic?ref=vettedconsumer.com and Simon Willison https://simonwillison.net/2026/Jul/16/kimi-k3/?ref=vettedconsumer.com had developers preferring K3 over every leading US model on that one narrow task. Willison also ran his usual "pelican on a bicycle" SVG doodle and got a valid result, while cautioning it is not a real capability test. The takeaway: a Chinese lab shipping an open-weight model that an independent index rates alongside Opus 4.8 is the actual story here, and it lands whether or not the individual self-reported cells hold up under scrutiny after July 27. The one spec that decides local speed is missing For a model you plan to run yourself, two numbers matter, and they do different jobs. Total parameters decide whether it fits in your memory. Active parameters decide how fast it decodes , because token generation is memory-bandwidth-bound: each token only reads the experts that fire, not all 2.8T weights. Our Mixture-of-Experts explainer https://vettedconsumer.com/mixture-of-experts-moe-explained-why-active-parameters-decide-what-runs-on-your-machine/ walks through why. Moonshot gave the routing "16 of 896 experts" but not a per-expert dimension, so the active-parameter count in billions cannot be derived from what is public. The prior generation, Kimi K2, was 1T total with 32B active. If K3 kept a similar active budget, it could decode at a usable clip despite its size; if the active count grew, it would be slower. We do not know, and any specific active-parameter figure circulating right now is an aggregator's guess. That single missing number is why a reliable tok/s estimate is impossible today. The hardware reality: nothing you own runs it Here is where the "largest open model ever" headline collides with a desk. The math below is projected from the confirmed 2.8T figure and the measured sizes of Kimi K2, whose real GGUF files ran about 587GB at Q4 and 1.09TB at Q8. Scale that by roughly 2.8x, and cross-check against the file-size rule bytes = parameters x bits-per-weight / 8 : | Quant | Projected K3 weights 2.8T | For comparison: Kimi K2 1T | |---|---|---| | Q8 | ~2.8 to 3.0 TB | ~1.09 TB | | Q4 K M | ~1.6 to 1.7 TB | ~587 GB | | Heavy IQ2 | ~0.9 to 1.05 TB | ~330 to 373 GB | Those are weights only. Add roughly 10 to 15 percent runtime overhead plus KV cache on top. The one mercy is that Kimi Delta Attention is a linear-attention scheme, so the KV/state footprint should stay modest even at long context on the order of tens of gigabytes at the full 1M window, not hundreds . With K3 the weights are the whole problem. Run those numbers against real machines and the result is stark: | Machine usable memory | Can it hold K3? | |---|---| | RTX 5090, 32GB | No, at any quant off by ~30x even for IQ2 | | 2x to 4x RTX 3090 48 to 96GB | No, at any quant | | Strix Halo, 128GB unified | No IQ2 needs ~0.9TB, about 8x short | | Mac Studio M3 Ultra, 512GB | No, even at the most brutal IQ2 ~0.9TB 512GB | | 8x H100 640GB | No, even for IQ2 | | 8x H200 1,128GB | IQ2 yes with room for KV; full Q4 no | The Mac Studio row is the headline reversal. A single 512GB Mac Studio M3 Ultra runs the current Kimi K2 at Q2 comfortably, around 380GB of weights with room to spare. The same machine cannot hold K3 at any quant. A 2.8x jump in size walked the ceiling right past the biggest single box a prosumer can buy. The smallest realistic way to self-host K3, once weights ship, is a small cluster: two or three Mac Studio 512GB machines linked with a distributed runtime EXO or MLX to reach 1 to 1.5TB of unified memory for an IQ2-to-low-Q3 quant, which is roughly a $20,000 to $30,000-plus proposition, or a single 8x H200 / B200-class server, which is rental-or-datacenter territory. Disk-offloading a 2.8T model through llama.cpp is technically possible and would run at fractions of a token per second, so it does not count as a real answer. For anyone at or below one 512GB Mac, a Strix Halo, a 5090, or a 3090 rig, K3 simply does not fit. So should you care? As a local model today, Kimi K3 is not one. It is an API model $3 per million input tokens, $15 per million output, per Constellation Research https://www.constellationr.com/insights/news/moonshot-ai-launches-kimi-k3?ref=vettedconsumer.com with an open-weight release promised for July 27, and even then it will be a cluster-only download that a rounding error of readers can host. If you want to use its intelligence now, the practical path is the API or renting a big GPU node by the hour, which is exactly the buy-vs-rent-vs-API question our cost calculator https://vettedconsumer.com/cost-calculator/ is built for. This is the same wall we hit with GLM-5.2 https://vettedconsumer.com/glm-5-2-the-most-powerful-open-weight-model-yet-and-the-brutal-reality-of-running-it-locally/ and Kimi K2.7 https://vettedconsumer.com/kimi-k2-7-code-trillion-parameter-open-coder-local-reality/ , only taller. Why it still matters: an open-weight model that an independent index rates next to Claude Opus 4.8 means the open frontier now touches the closed one, and MXFP4-native training is a signal that these labs are designing for the hardware reality, not ignoring it. The things to watch after July 27 are the actual active-parameter count which will tell us how fast it can possibly decode and the first community quants, since a clever IQ2 or a smaller distilled variant is the only path that ever puts a model this size on a machine you own. When the weights and the technical report land, we will update this with measured numbers. Sources and how we researched this - Announcement, specs, pricing, and self-reported benchmarks: Moonshot AI's Kimi K3 blog https://www.kimi.com/blog/kimi-k3?ref=vettedconsumer.com and API docs https://platform.kimi.ai/docs/guide/kimi-k3-quickstart?ref=vettedconsumer.com primary, vendor-stated . - Independent evaluation: Artificial Analysis https://artificialanalysis.ai/models/kimi-k3?ref=vettedconsumer.com Intelligence Index 57, ranked 3rd to 4th . - Reporting and context: Constellation Research https://www.constellationr.com/insights/news/moonshot-ai-launches-kimi-k3?ref=vettedconsumer.com , Axios https://www.axios.com/2026/07/16/moonshot-kimi-ai-china-model-openai-anthropic?ref=vettedconsumer.com , VentureBeat https://venturebeat.com/technology/chinas-moonshot-ai-releases-kimi-k3-the-largest-open-source-model-ever-rivaling-top-u-s-systems?ref=vettedconsumer.com , and Simon Willison https://simonwillison.net/2026/Jul/16/kimi-k3/?ref=vettedconsumer.com . - Weights-not-yet-out status: confirmed by the blog "by July 27, 2026" and by Moonshot's GitHub org https://github.com/MoonshotAI?ref=vettedconsumer.com , which shows no Kimi-K3 repository as of the announcement. - Hardware footprint: projected from the measured Kimi K2 GGUF sizes and the file-size rule; not measured on K3. We have not run the model first-hand. Related: Kimi K2.7 Code and the 594GB reality · Mac Studio M3 Ultra for local AI · How much VRAM a 70B needs · GGUF quantization, explained