- Size
- 2.8T parametersfrom the GGUF metadata of a conversion, because this repo's own weights are packed: unsloth/Kimi-K3-GGUF
- Licence
- unknown, Kimi K3 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" 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
- 13 Jun 2026the date the repo was created
- Context length
- 1M tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)KimiK3ForConditionalGeneration
- On Hugging Face
- moonshotai/Kimi-K3734k downloads in the last 30 days, 12k likes
Scores
- effort high37.0%±4.913 Oct 2026, log
- effort max36.0%±4.813 Oct 2026, log
- effort low18.0%±3.913 Oct 2026, log
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 79.0% (Anthropic). All 166 results
- effort max0.0%28 Aug 2026, Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: gpt-6.2-astra_max, 10.3% (OpenAI). All 16 results
- effort max39.0%±7.717 Jul 2026, log
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 100.0% (Anthropic). All 83 results
- effort max72.2%±2.717 Jul 2026, Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 95.1% (Anthropic). All 127 results
- effort max34.2%±6.111 Sep 2026, Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 90.0% (Anthropic). All 41 results
- effort max93.1%±1.516 Jul 2026, log
- effort high91.9%±1.97 Aug 2026, log
- effort low84.8%±2.67 Aug 2026, log
Epoch's name: kimi-k3. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- effort max54.4%±11.43 Sep 2026, log
Epoch's name: kimi-k3. Best closed model on this set: claude-sonnet-5-5_max, 97.8% (Anthropic). All 47 results
Epoch's name: kimi-k3. Best closed model on this set: gpt-6.2-astra_max, 95.0% (OpenAI). All 148 results
- effort max97.2%±1.116 Jul 2026, log
- effort high93.3%±3.87 Aug 2026, log
- effort low68.9%±7.07 Aug 2026, log
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- effort max50.6%±1.627 Aug 2026, log
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 94.0% (Anthropic). All 105 results
- effort unknown47.8%via Epoch AI (error bar left out)
Epoch's name: kimi-k3. Best closed model on this set: gemini-4-carbon_unknown, 84.2% (Google DeepMind). All 66 results
- effort max94.5%arcprize.org
- effort high86.7%arcprize.org
- effort low65.7%arcprize.org
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5_max, 98.5% (Anthropic). All 318 results
- effort max60.4%via Epoch AI
- effort high55.0%via Epoch AI
- effort low12.4%via Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: gpt-6-astra_max, 95.0% (OpenAI). All 299 results
- effort unknown29.5%andonlabs.com
Epoch's name: kimi-k3. Best closed model on this set: gemini-4-carbon_unknown, 58.6% (Google DeepMind). All 38 results
- effort max23.4%via Epoch AI
- effort low3.1%via Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- effort max; Harness: mini-swe-agent; Reasoning effort: max68.5%64.0 to 73.1deepswe.datacurve.ai
Epoch's name: kimi-k3. Best closed model on this set: gpt-6-astra_xhigh, 74.1% (OpenAI). All 69 results
- effort max91.2%conceptualreasoning.ai
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- effort unknown; Harness: mini-swe-agent; Reasoning effort: none44.2%cognition.com
Epoch's name: kimi-k3. Best closed model on this set: claude-opus-5-5_medium, 54.6% (Anthropic). All 46 results
- effort max; Harness: proximus25.9%frontierswe.com
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_max, 69.9% (Anthropic). All 25 results
- effort unknown48.3%20 Jul 2026, gbaeval.com
Epoch's name: kimi-k3. Best closed model on this set: claude-opus-5-5_unknown, 90.6% (Anthropic). All 28 results
- effort max39.8%surgehq.ai
Epoch's name: kimi-k3. Best closed model on this set: gemini-4-carbon_high, 56.4% (Google DeepMind). All 19 results
- effort max52.7%49.0 to 56.4conceptualreasoning.ai
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results
- effort unknown; Scaffold: Claude Code31.5%via Epoch AI
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-1_max, 44.6% (Anthropic). All 22 results
- effort unknown87.0%±3.4vals.ai
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-1_max, 100.0% (Anthropic). All 85 results
- effort max60.7%lmcouncil.ai
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5-2_unknown, 94.9% (Anthropic). All 125 results
- effort unknown95.0%yhenon.github.io
- effort max93.0%yhenon.github.io
Epoch's name: kimi-k3. Best closed model on this set: claude-fable-5_high, 95.0% (Anthropic). All 29 results
- effort max82.6%±0.0htihle.github.io
Epoch's name: kimi-k3. Best closed model on this set: gpt-6-astra_promax, 93.6% (OpenAI). All 173 results
- effort high; Agent: kimi_code; Reasoning effort: high15.4%htihle.github.io
Epoch's name: kimi-k3. 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), 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
unsloth/Kimi-K3-GGUF (converted by unsloth), 193k 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 |
|---|---|---|---|
| UD-Q1_0 | 434.3 GBfile size, 11 parts | 435.7 GB434.3 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q1_0Download hf download unsloth/Kimi-K3-GGUF --include "UD-Q1_0/Kimi-K3-UD-Q1_0-*-of-00011.gguf" --local-dir Kimi-K3-GGUF |
| UD-TQ1_0 | 474.0 GBfile size, 12 parts | 475.3 GB474.0 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-TQ1_0Download hf download unsloth/Kimi-K3-GGUF --include "UD-TQ1_0/Kimi-K3-UD-TQ1_0-*-of-00012.gguf" --local-dir Kimi-K3-GGUF |
| UD-TQ2_0 | 513.6 GBfile size, 13 parts | 514.9 GB513.6 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-TQ2_0Download hf download unsloth/Kimi-K3-GGUF --include "UD-TQ2_0/Kimi-K3-UD-TQ2_0-*-of-00013.gguf" --local-dir Kimi-K3-GGUF |
| UD-IQ1_S | 553.2 GBfile size, 14 parts | 554.6 GB553.2 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-IQ1_SDownload hf download unsloth/Kimi-K3-GGUF --include "UD-IQ1_S/Kimi-K3-UD-IQ1_S-*-of-00014.gguf" --local-dir Kimi-K3-GGUF |
| UD-IQ1_M | 604.3 GBfile size, 15 parts | 605.6 GB604.3 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-IQ1_MDownload hf download unsloth/Kimi-K3-GGUF --include "UD-IQ1_M/Kimi-K3-UD-IQ1_M-*-of-00015.gguf" --local-dir Kimi-K3-GGUF |
| UD-IQ2_XXS | 662.2 GBfile size, 16 parts | 663.6 GB662.2 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-IQ2_XXSDownload hf download unsloth/Kimi-K3-GGUF --include "UD-IQ2_XXS/Kimi-K3-UD-IQ2_XXS-*-of-00016.gguf" --local-dir Kimi-K3-GGUF |
| UD-Q2_K_XL | 802.1 GBfile size, 19 parts | 803.4 GB802.1 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q2_K_XLDownload hf download unsloth/Kimi-K3-GGUF --include "UD-Q2_K_XL/Kimi-K3-UD-Q2_K_XL-*-of-00019.gguf" --local-dir Kimi-K3-GGUF |
| UD-Q4_K_XL | 1405.1 GBfile size, 32 parts | 1406.4 GB1405.1 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XLDownload hf download unsloth/Kimi-K3-GGUF --include "UD-Q4_K_XL/Kimi-K3-UD-Q4_K_XL-*-of-00032.gguf" --local-dir Kimi-K3-GGUF |
| UD-Q8_K_XL | 1453.9 GBfile size, 34 parts | 1455.3 GB1453.9 + 0.82 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q8_K_XLDownload hf download unsloth/Kimi-K3-GGUF --include "UD-Q8_K_XL/Kimi-K3-UD-Q8_K_XL-*-of-00034.gguf" --local-dir Kimi-K3-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX mxfp4pipenetwork/Kimi-K3-REAP80-MLX-mxfp4-q8 | 325.7 GBfile size | 327.0 GB325.7 + 0.82 + 0.50 | Download page |
| MLX mxfp4pipenetwork/Kimi-K3-REAP73-MLX-mxfp4-q8 | 420.4 GBfile size | 421.7 GB420.4 + 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 --local-dir Kimi-K3Other conversions (16; sizes not read)
- nvidia/Kimi-K3-NVFP4 NVFP4, 3.3k downloads
- nikita-savelyev-cerebras/tiny-random-kimi-k3 Other, 2.4k downloads
- nikita-savelyev-cerebras/tiny-random-kimi-k3-mxfp4 Other, mxfp4, 1.4k downloads
- RedHatAI/Kimi-K3-NVFP4 NVFP4, 988 downloads
- tiny-random/kimi-k3 Other, 634 downloads
- unsloth/Kimi-K3 Other, 571 downloads
- ISTA-DASLab/Kimi-K3-P48NVFP4-MoESQ NVFP4, 494 downloads
- RedHatAI/Kimi-K3-FP8-BLOCK FP8, 346 downloads
- GrEarl/Kimi-K3-NVFP4A16-Requantized NVFP4, 264 downloads
- open-inference/Kimi-K3-W4A8-K512-DenseChannel-GPTQ-Cal128-v2 GPTQ, 222 downloads
- vessl/Kimi-K3-W4AFP8 FP8, 184 downloads
- wrldsuksgo2mars/Kimi-K3-EXL3-K2-Core-K6 EXL3, 88 downloads
- Radioheading/Kimi-K3-W4AFP8 FP8, 64 downloads
- wrldsuksgo2mars/Kimi-K3-EXL3-K2 EXL3, 59 downloads
- cwnd/Kimi-K3-MoonViT-V2-INT8-W8A16 bitsandbytes, int8, 0 downloads
- Zhangdanyang/Kimi-K3-DSpark-FP8-PTPC FP8, 0 downloads
The Hub tags 57 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
18 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 |
|---|---|---|---|---|---|---|---|---|---|
| Wafer | $0.28 | $9.60 | 1M | mxfp4 | 90 | 0.6 s | 100.0% | yes | yes |
| Relace | $0.18 | $11 | 1M | fp4 | 69 | 1.1 s | 99.9% | yes | no |
| DigitalOcean | $2.55 | $12.9 | 1M | not given | 29 | 1.7 s | 100.0% | yes | yes |
| Io Net | $1.90 | $14 | 1024k | fp4 | 32 | 2.9 s | 99.3% | yes | yes |
| DeepInfra | $2.85 | $14.3 | 1M | mxfp4 | 58 | 0.9 s | 99.8% | yes | yes |
| Amazon Bedrock | $3.00 | $15 | 1M | not given | 95 | 2.6 s | 99.9% | yes | no |
| BaseTen | $3.00 | $15 | 1M | fp8 | 55 | 1.1 s | 99.6% | yes | no |
| Fireworks | $3.00 | $15 | 1M | not given | 39 | 1.6 s | 99.7% | yes | yes |
| InferenceNet | $0.24 | $15 | 1M | fp4 | 80 | 0.5 s | 100.0% | yes | yes |
| Modal | $3.00 | $15 | 1M | mxfp4 | 67 | 0.8 s | 100.0% | yes | yes |
| Moonshot AIthe maker | $3.00 | $15 | 1M | not given | 36 | 3.4 s | 99.4% | yes | yes |
| Nebius | $3.00 | $15 | 1M | fp4 | 115 | 1.8 s | 100.0% | yes | yes |
| Parasail | $3.00 | $15 | 1M | fp4 | 26 | 1.2 s | 100.0% | yes | yes |
| River | $3.00 | $15 | 1M | not given | 124 | 0.3 s | 99.6% | no | yes |
| Together | $3.00 | $15 | 1M | not given | 62 | 1.1 s | 100.0% | yes | yes |
| Sail Research | $0.34 | $16.6 | 1M | fp4 | 58 | 0.7 s | 100.0% | yes | yes |
| Makora | $1.00 | $18 | 1M | fp4 | 36 | 1.4 s | 99.9% | yes | yes |
| featherless-ai | ? | ? | ? | not given | ? | ? | ? | not given | not given |
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:wafer", "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.
- Inferact/Kimi-K3-DSpark3.6B11kUnreviewed
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.