Smaller alternatives
Models that need less memory than Kimi Linear 48B A3B and score about as well or better.
Kimi Linear 48B A3B
MoE
Moonshot AI · 49.1B (3B active) · Mixture of experts
Kimi Linear 48B A3B is a 49.1B model from Moonshot AI with a 1M-token context. At Q4_K_M with an 8K context it needs about 28.4 GB; 37 GB leaves room for longer chats.
Mixture of experts
Parameters: 49.1B · Active: 3B
All 49.1B parameters load into memory, but only 3B work on each token, so it runs at the speed of a much smaller model.
Quantization options
| Quant | Bits | Memory | Quality | On your hardware |
|---|---|---|---|---|
| Q2_K | 3.16 | — | Noticeably worse | … |
| Q3_K_M | 4 | — | Some loss | … |
| Q4_K_M | 4.89 | — | Good | … |
| Q5_K_M | 5.7 | — | Good | … |
| Q6_K | 6.56 | — | Near original | … |
| Q8_0 | 8.5 | — | Near original | … |
| F16 | 16 | — | Original | … |
Memory by context length
At Q4_K_M. The context cache grows with every token the model keeps in mind.
| Context | Context cache | Total |
|---|---|---|
| 4K | 0 GB | 28.3 GB |
| 8K | 0.1 GB | 28.4 GB |
| 32K | 0.2 GB | 28.5 GB |
| 128K | 1 GB | 29.3 GB |
| 256K | 2 GB | 30.3 GB |
Next steps
Compare it with
Scored models of a similar size, side by side on your device.
Least hardware that runs it well
Desktop devices with the least memory that give Kimi Linear 48B A3B grade A or S at Q4_K_M.
Can I run Kimi Linear 48B A3B locally?
Can I run Kimi Linear 48B A3B locally?
Yes, if your GPU or Mac has about 28.4 GB free for it at Q4_K_M. Open this page on that computer to see the grade for your exact hardware, then start it with llama.cpp, Ollama or LM Studio.
How much memory does Kimi Linear 48B A3B need?
About 28.4 GB at Q4_K_M with an 8K context and 49 GB at Q8_0. Lower quantizations fit smaller cards with some loss in quality; longer contexts add to the total.