Kimi-K3
Kimi-K3 is an open-weight vision and language model from Moonshot AI with 2.78T parameters (mixture of experts). At 4-bit it needs about 1858.6 GB of memory, more than any single device in our directory has.
Overview
Includes room for an 8k context and runtime overhead. More context needs more memory.
Key information
- Lab
- Moonshot AI
- Released
- 13 Jun 2026
- Parameters
- 2.78T · MoE
- Type
- Vision and language
- Context
- 1,048,576 tokens
- License
- Custom license
- HF downloads, 30 days
- 790,518
- Repository
- moonshotai/Kimi-K3
More from Moonshot AI
View allWill it run on your hardware?
How we estimate| Estimate or device | Memory |
|---|---|
| Q4 | 1858.6 GB |
| Q8 | 3265.2 GB |
| BF16 | 6139.7 GB |
| RTX 5060 Ti 16 GB | 16 GB usable |
| RTX 5090 32 GB | 32 GB usable |
| DGX Spark 128 GB | 108.8 GB usable |
| Mac Studio M5 Ultra 256 GB | 192 GB usable |
Single-device memory estimate. Pick a device to update the On your device column.
36 GB usable memory
| Format | Memory needed | On your device | Example hardware |
|---|---|---|---|
| 4-bit (Q4) | ≈ 1858.6 GB | Too big 5163% | No single device in our directory has enough memory |
| 8-bit (Q8) | ≈ 3265.2 GB | Too big 9070% | No single device in our directory has enough memory |
| 16-bit (BF16) | ≈ 6139.7 GB | Too big 17055% | No single device in our directory has enough memory |
Fits means the estimate stays within 90% of usable memory. Tight is above 90% and up to 100%. Too big is above 100%. These are memory checks, not speed or compatibility tests.
Mac allowances are 66% of total RAM up to 36 GB, and 75% above. AI boxes allow 85% because their memory is shared with the operating system.
Where to run it
With GGUF
Choose a file for llama.cpp or a compatible app.
unsloth/Kimi-K3-GGUFFind compatible hardwareThrough an API
Wafer (mxfp4) is the cheapest provider we found.
$0.276 / $9.60
Input / output per 1M tokens
₹26 / ₹905 at ₹94.29 per $1
Blended: $2.607 (₹246) per 1M. Ranked by a 3:1 mix of input and output tokens.
llama-cli -hf unsloth/Kimi-K3-GGUF \
-c 8192 -p "Explain how a transformer works."Example commands built from the repository ids above. They need a working install of llama.cpp. We have not run them; check each repository for available files.
| Precision | |||||
|---|---|---|---|---|---|
| WaferCheapest | $2.607₹246 | $0.276₹26 | $9.60₹905 | 1,048,576 | MXFP4 |
| Relace | $2.8882₹272 | $0.1824₹17 | $11.0055₹1,038 | 1,048,576 | FP4 |
| Chutes | $3.7625₹355 | $1.85₹174 | $9.50₹896 | 1,048,576 | MXFP4 |
| InferenceNet | $3.9255₹370 | $0.234₹22 | $15.00₹1,414 | 1,048,576 | FP4 |
| Sail Research | $4.4037₹415 | $0.345₹33 | $16.58₹1,563 | 1,048,576 | FP4 |
| Makora | $4.90₹462 | $1.20₹113 | $16.00₹1,509 | 1,048,576 | FP4 |
| DigitalOcean | $5.15₹486 | $2.55₹240 | $12.95₹1,221 | 1,048,576 | Not listed |
| Io Net | $5.225₹493 | $2.30₹217 | $14.00₹1,320 | 1,048,320 | FP4 |
| DeepInfra | $5.70₹537 | $2.85₹269 | $14.25₹1,344 | 1,048,576 | MXFP4 |
| Amazon Bedrock | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| BaseTen | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | FP8 |
| Fireworks | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| KT | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| Modal | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | MXFP4 |
| Moonshot AI | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| Nebius | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | FP4 |
| Parasail | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | FP4 |
| River | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| Together | $6.00₹566 | $3.00₹283 | $15.00₹1,414 | 1,048,576 | Not listed |
| Alibaba | $6.90₹651 | $3.45₹325 | $17.25₹1,627 | 1,048,576 | Not listed |
INR at ₹94.29 per $1. Prices fetched 14 Nov 2026.
More about this model
Frequently asked questions
How much memory does Kimi-K3 need? At an 8k context, about 1858.6 GB at 4-bit, 3265.2 GB at 8-bit and 6139.7 GB at 16-bit (BF16). A longer context needs more.
Can I run Kimi-K3 on a Mac? No Mac in our directory has enough usable memory for the 4-bit estimate of 1858.6 GB. We have not found a trusted MLX build yet, but a GGUF build is available.
Does the full context fit? Kimi-K3 supports up to 1,048,576 tokens of context. Our memory estimates use 8k, so long prompts need more memory than shown.
What does it cost through an API? The cheapest provider we found is Wafer, at $0.28 per 1M input tokens and $9.60 per 1M output tokens (₹26 and ₹905 at 94.29 rupees per dollar). Ranked by a 3:1 mix of input and output tokens.
License in plain English
Custom license. This model uses a custom license rather than a standard one. Read it on the model card before using the model commercially. This summary is not legal advice.
OMHQ RigBench
Not yet tested. We show no speed or quality score without a measurement.
Revision history and sources
Last updated . Sources:Hugging Face model card, GGUF repository, OpenRouter.