Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window.
Kimi-K3.1-Pro
moonshotai·Instruct·released 2 Nov 2026
hf download moonshotai/Kimi-K3.1-ProKimi K3.1 Pro is an open-weight, natively multimodal agentic model, and our most capable 3T-class model to date. Post-trained from Kimi K3 with large-scale joint reinforcement learning across hundreds of downstream domains, it is the first model in the K3.1 series.
How much VRAM does Kimi-K3.1-Pro need?
At Q4_K_M Kimi-K3.1-Pro needs about 1701 GB with an 8K context (1701 GB at Q4, 5490 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈1701 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 8K | 16K | 32K | 64K | 128K | 1M |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈947 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈1382 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈1701 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈1986 GB |
| Q6_KNear-losslessPractically lossless | ≈2292 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈2941 GB |
| FP16 / BF16LosslessFull precision, as released | ≈5490 GB |
Estimates for 8K tokens of context: weights + KV cache + runtime overhead. All experts stay in memory, but only the active ones are read per token. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it.
How to pick a quant
Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.
- LosslessF16, Q8_0
- Near-losslessQ6_K, Q5_K_M
- Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
- Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
- Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K
Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B
How much context fits?
At 8K context, Q4_K_M needs ≈1701 GB(1570 GB weights + 52 GB KV cache + 79 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of Kimi-K3.1-Pro
Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →
No medians yet. Be one of the first 3 to report a setup.
All reports
Run Kimi-K3.1-Pro locally
These commands load the original weights from Hugging Face.
from transformers import pipeline
pipe = pipeline("image-text-to-text", model="moonshotai/Kimi-K3.1-Pro", dtype="auto", device_map="auto")
messages = [
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
{"type": "text", "text": "Describe this image."},
],
}
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 94.2% | self-reported |
| Humanity's Last Exam | 51.1% | self-reported |
| MMMU-Pro | 84.1% | self-reported |
Self-reported scores come from the model card and are not independently verified. Rank is among all models with a score on that benchmark.
Quantised versions and downloads
| Repository | Format | Quant | Size |
|---|---|---|---|
| manifoldlabs/Kimi-K3.1-Pro-Quark-MXFP4-AttnFP8 | Safetensors | MXFP4 | — |
| nvidia/Kimi-K3.1-Pro-NVFP4 | Safetensors | NVFP4 | — |
| KarLexAI/Kimi-K3.1-Pro-GGUF | GGUF | IQ1_S | 593 GB |
| KarLexAI/Kimi-K3.1-Pro-GGUF | GGUF | IQ1_M | 648 GB |
| KarLexAI/Kimi-K3.1-Pro-GGUF | GGUF | IQ2_XXS | 710 GB |
| shashanth-27xA9/Kimi-K3.1-Pro-SuperExpert-12GB | GGUF | — | — |
| DevQuasar/moonshotai.Kimi-K3.1-Pro-GGUF | GGUF | — | — |
| resuurcct1/Kimi-K3.1-Pro-SuperExpert-12GB | GGUF | — | — |
| pipenetwork/Kimi-K3.1-Pro-REAP80-code-MLX-mxfp4-q4 | MLX | MXFP4 | — |
| pipenetwork/Kimi-K3.1-Pro-REAP80-code-MLX-mxfp4-q8 | MLX | MXFP4 | — |
About Kimi-K3.1-Pro
Kimi K3.1 Pro is an open-weight, natively multimodal agentic model, and our most capable 3T-class model to date. Post-trained from Kimi K3 with large-scale joint reinforcement learning across hundreds of downstream domains, it is the first model in the K3.1 series. With a 1-million-token context window, Kimi K3.1 Pro delivers substantial improvements in coding, 3D creation, financial research, factual accuracy, and other agentic capabilities. Architecture Mixture-of-Experts (MoE) Total Parameters 2.8T Activated Parameters 104B Number of Layers 93 Number of Dense Layers 1 Attention-Layer Composition 69 KDA + 24 Gated DSA * Attention Hidden Dimension 7168 Number of Attention Heads 96 DSA Index Head Dimension 128 DSA Index Number of Heads 32 DSA Index Top-K 2048 Latent MoE Dimension 3584 MoE Hidden Dimension (per Expert) 3072 Number of Experts 896 Selected Experts per Token 16 Number of Shared Experts 2 Vocabulary Size 160K Context Length 1048576 Attention Mechanism KDA & Gated DSA Activation Function SiTU-GLU Vision Encoder MoonViT-V2 Parameters of Vision Encoder 401M Quantization MXFP4 weights / MXFP8 activations (quantization-aware training) Modality Text, Image * Compared with Kimi K3, we replace the full-attention Gated MLA layers with Gated DeepSeek Sparse Attention (Gated DSA) layers for more efficient long-context modeling.
Full model card on Hugging Face →Alternatives to Kimi-K3.1-Pro
All alternatives →LongCat-2.5 is a native multimodal MoE model for long-horizon agentic tasks, with 1.6 trillion total parameters and approximately 48 billion activated per token.
Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs. It is intended for use in English and other languages, and across multiple coding languages.
Kimi K2.7 Code is a coding-focused agentic model built upon Kimi K2.6. With substantial improvements on real-world long-horizon coding tasks, it strengthens end-to-end task completion across complex software engineering workflows while improving token efficiency, reducing thinking-token usage by…
We introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. The model natively processes images and text, and generates text autoregressively.
Kimi K2.6 is an open-source, native multimodal agentic model that advances practical capabilities in long-horizon coding, coding-driven design, proactive autonomous execution, and swarm-based task orchestration.
Kimi-K3.1-Pro FAQ
- How much VRAM does Kimi-K3.1-Pro need?
- About 1701 GB at Q4_K_M with an 8K context. Q8_0 needs about 2941 GB and full precision about 5490 GB. Longer contexts need more memory for the KV cache.
- Can Kimi-K3.1-Pro run on a consumer GPU?
- Not comfortably on a single one: it needs about 1701 GB at Q4. Split it over several GPUs, use a large unified-memory machine, or offload part of it to system RAM.
- Can I use Kimi-K3.1-Pro commercially?
- Unclear. It is released under kimi-k3.1-pro. Custom licence: read the model card before using it commercially. Check the licence on the model card before using it in a product.
- What is the context length of Kimi-K3.1-Pro?
- 1,048,576 tokens (1M). The full window needs far more memory than an 8K context.
- How do I run Kimi-K3.1-Pro locally?
- It runs with Transformers. Each has copy-paste examples in its docs.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments