- Size
- 2.8T parametersfrom the GGUF metadata of a conversion, because this repo's own weights are packed: KarLexAI/Kimi-K3.1-Pro-GGUF
- Licence
- unknown, Kimi K3.1-Pro LicenseOur reading, check the licence: Commercial use is allowed, but if you or an affiliate run a Model as a Service business and your combined revenue exceeds 20 million US dollars over any consecutive 12 months, you need a separate agreement with Moonshot AI before commercial use. Products or services with more than 100 million monthly active users or more than 20 million US dollars in monthly revenue must prominently display "Kimi K3.1-Pro" in their user interface. Neither rule applies to internal use or to use through Moonshot AI's official products or certified inference partners; otherwise it is MIT style, with no duty to share changes.
- On the Hub since
- 2 Nov 2026the date the repo was created
- Context length
- 1M tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)KimiK31ForConditionalGeneration
- On Hugging Face
- moonshotai/Kimi-K3.1-Pro61k downloads in the last 30 days, 2.3k likes
Scores
- effort max45.0%±5.06 Nov 2026, log
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5-2_max, 79.0% (Anthropic). All 166 results
- effort max65.9%±7.56 Nov 2026, Epoch AI
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5-2_max, 100.0% (Anthropic). All 83 results
- effort max78.2%±2.46 Nov 2026, Epoch AI
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5-2_max, 95.1% (Anthropic). All 127 results
- effort max55.8%±6.46 Nov 2026, Epoch AI
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5-2_max, 90.0% (Anthropic). All 41 results
- effort max94.2%±1.46 Nov 2026, log
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- effort max42.0%±5.06 Nov 2026, Epoch AI
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: gpt-6.2-astra_max, 95.0% (OpenAI). All 148 results
- effort max100.0%6 Nov 2026, log
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- effort max62.2%±1.56 Nov 2026, log
Epoch's name: kimi-k3.1-pro-together. Best closed model on this set: claude-fable-5-2_max, 94.0% (Anthropic). All 105 results
- effort unknown65.5%via Epoch AI (error bar left out)
Epoch's name: kimi-k3.1-pro. Best closed model on this set: gemini-4-carbon_unknown, 84.2% (Google DeepMind). All 66 results
- effort unknown24.6%via Epoch AI
Epoch's name: kimi-k3.1-pro. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- effort max93.3%conceptualreasoning.ai
Epoch's name: kimi-k3.1-pro. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- effort max46.6%surgehq.ai
Epoch's name: kimi-k3.1-pro. Best closed model on this set: gemini-4-carbon_high, 56.4% (Google DeepMind). All 19 results
- effort max57.6%54.2 to 61.0conceptualreasoning.ai
Epoch's name: kimi-k3.1-pro. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results
- effort unknown91.0%±2.9vals.ai
Epoch's name: kimi-k3.1-pro. Best closed model on this set: claude-fable-5-1_max, 100.0% (Anthropic). All 85 results
- effort xhigh; Agent: kimi_code; Reasoning effort: xhigh18.9%htihle.github.io
Epoch's name: kimi-k3.1-pro. Best closed model on this set: claude-fable-5-2_xhigh, 65.3% (Anthropic). All 27 results
Claimed by the maker Claimed
The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.
Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.
Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.82 GB, worked out from the model's config (moonshotai/Kimi-K3.1-Pro), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
GGUF for llama.cpp, Ollama, LM Studio and others
KarLexAI/Kimi-K3.1-Pro-GGUF (converted by KarLexAI), 9.9k downloads. Sizes are the real files on the Hub, all parts added up. For image input you also need the projector file mmproj-F16.gguf (0.84 GB), not counted below.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| IQ1_S | 552.0 GBfile size, 14 parts | 553.3 GB552.0 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf KarLexAI/Kimi-K3.1-Pro-GGUF:IQ1_SDownload hf download KarLexAI/Kimi-K3.1-Pro-GGUF --include "KLX-IQ1_S/Kimi-K3.1-Pro-KLX-IQ1_S-*-of-00014.gguf" --local-dir Kimi-K3.1-Pro-GGUF |
| IQ1_M | 603.4 GBfile size, 15 parts | 604.7 GB603.4 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf KarLexAI/Kimi-K3.1-Pro-GGUF:IQ1_MDownload hf download KarLexAI/Kimi-K3.1-Pro-GGUF --include "KLX-IQ1_M/Kimi-K3.1-Pro-KLX-IQ1_M-*-of-00015.gguf" --local-dir Kimi-K3.1-Pro-GGUF |
| IQ2_XXS | 661.1 GBfile size, 17 parts | 662.4 GB661.1 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf KarLexAI/Kimi-K3.1-Pro-GGUF:IQ2_XXSDownload hf download KarLexAI/Kimi-K3.1-Pro-GGUF --include "KLX-IQ2_XXS/Kimi-K3.1-Pro-KLX-IQ2_XXS-*-of-00017.gguf" --local-dir Kimi-K3.1-Pro-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX mxfp4pipenetwork/Kimi-K3.1-Pro-REAP80-code-MLX-mxfp4-q4 | 299.8 GBfile size | 301.1 GB299.8 + 0.82 + 0.50 | Download page |
| MLXinferencerlabs/Kimi-K3.1-Pro-MLXse | 1454.0 GBfile size | 1455.3 GB1454.0 + 0.82 + 0.50 | Download page |
Original weights
The maker's own weights are packed into integer tensors, so the safetensors header gives no size we can trust; the repo lists the real file sizes.
hf download moonshotai/Kimi-K3.1-Pro --local-dir Kimi-K3.1-ProOther conversions (10; sizes not read)
- local-inference-lab/Kimi-K3.1-Pro-EXL3-2bit EXL3, 2bit, 62k downloads
- shashanth-27xA9/Kimi-K3.1-Pro-SuperExpert-12GB GGUF, 9.6k downloads; sizes not used: every file is far too small for this model; the repo may be tagged wrongly or hold only part of it
- nvidia/Kimi-K3.1-Pro-NVFP4 NVFP4, 2.2k downloads
- tiny-random/kimi-k3.1-pro Other, 655 downloads
- dphn/Kimi-K3.1-Pro-NF12 Other, 89 downloads
- manifoldlabs/Kimi-K3.1-Pro-Quark-MXFP4-AttnFP8 FP8, mxfp4, 55 downloads
- technicallyAnnoyed/kimi-k3.1-pro-mini Other, 54 downloads
- wrldsuksgo2mars/Kimi-K3.1-Pro-EXL3-K2-MC EXL3, 54 downloads
- wrldsuksgo2mars/Kimi-K3.1-Pro-EXL3-K2-Core-K6 EXL3, 45 downloads
- tomcupr/Kimi-K3.1-Pro-DGX-Station-2bit Other, 2bit, 0 downloads
The Hub tags 23 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
19 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.
| Provider | In | Out | Context | Precision | Tokens/s | First token | Uptime 24h | Tools | JSON |
|---|---|---|---|---|---|---|---|---|---|
| Decart | $1.20 | $6.00 | 1M | mxfp4 | 41 | 3.2 s | 100.0% | yes | yes |
| Relace | $0.24 | $8.40 | 1M | fp4 | 66 | 1.5 s | 100.0% | yes | no |
| Wafer | $0.25 | $8.80 | 1M | mxfp4 | 92 | 0.6 s | 100.0% | yes | yes |
| InferenceNet | $1.80 | $9.00 | 1M | mxfp4 | 102 | 0.6 s | 100.0% | yes | yes |
| Sail Research | $1.04 | $9.35 | 1M | fp4 | 39 | 0.7 s | 100.0% | yes | yes |
| Makora | $2.00 | $9.49 | 1M | not given | 76 | 1.1 s | 97.3% | yes | yes |
| DeepInfra | $1.90 | $9.50 | 1M | bf16 | 41 | 0.9 s | 99.7% | yes | yes |
| AkashML | $1.75 | $10 | 1M | fp4 | 74 | 1.5 s | 99.8% | yes | yes |
| DigitalOcean | $2.00 | $10 | 1M | not given | 95 | 1.5 s | 100.0% | yes | yes |
| Fireworks | $2.00 | $10 | 1M | not given | 40 | 4.0 s | 97.5% | yes | yes |
| Modal | $2.00 | $10 | 1M | mxfp4 | 82 | 0.8 s | 100.0% | yes | yes |
| Moonshot AIthe maker | $2.00 | $10 | 1M | mxfp4 | 31 | 3.1 s | 99.9% | yes | yes |
| Nebius | $2.00 | $10 | 1M | fp4 | 164 | 0.8 s | 100.0% | yes | yes |
| Parasail | $2.00 | $10 | 1M | fp4 | 93 | 1.0 s | 99.9% | yes | yes |
| PrimaLabs | $2.00 | $10 | 1M | mxfp4 | 85 | 1.1 s | 99.8% | yes | yes |
| River | $2.00 | $10 | 1M | not given | 96 | 0.7 s | 99.9% | yes | yes |
| Together | $2.00 | $10 | 1M | not given | 40 | 1.6 s | 99.3% | yes | yes |
| Morph | $1.05 | $15.7 | 1M | fp8 | 43 | 2.2 s | 99.1% | yes | yes |
| Amazon Bedrock | $4.00 | $20 | 1M | not given | 117 | 1.7 s | 98.2% | yes | no |
Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.
curl https://router.huggingface.co/v1/chat/completions \
-H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \
-d '{"model": "moonshotai/Kimi-K3.1-Pro:decart", "messages": [{"role": "user", "content": "Hello"}]}'Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.
Fine-tunes of this model
Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.
- RadixArk/Kimi-K3.1-Dflash2-Preview2.8B7.7kUnreviewed
- Grawnch/Kimi-K3.1-Pro-Coder1.1kUnreviewed
- audnai/penclaw-Kimi-K3.1-pro-unrestricted786Unreviewed
- Blackfrost-Research/Kimi-K3.1-Pro-DERISKED63Unreviewed
- lyber-ai/Kimi-K3.1-Pro-UNCENSORED-MXFP425Unreviewed
Planned for this page Not built yet
- Try it in private chat
- A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
- Fine-tune it
- Fine-tuning a few base models from this page and publishing the result here.
- Creators and earnings
- Pages for people who publish fine-tunes and adapters of this model, and what they earn.
- Fund and bounties
- Bounties for new open models, paid from a public fund. How the fund will work.
- Contribute compute
- How to give compute to open model work, with Akash.
Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.