What GPU do I need to run moonshotai/Kimi-Linear-48B-A3B-Instruct?
49.1B parameters, published in BF16. View on Hugging Face
Kimi-Linear-48B-A3B-Instruct is published by moonshotai on Hugging Face, with 184,990 downloads and 589 likes to date. It's a KimiLinearForCausalLM model built for text-generation, published natively in BF16.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 91.5 GB | 109.8 GB | RTX 3090 | 5 | $0.880/hr |
| FP8 (quantized) | 45.7 GB | 54.9 GB | RTX PRO 6000 | 1 | $1.38/hr |
| cheaper alt. | RTX 5060 Ti | 4 | $0.581/hr | ||
| INT4 (quantized) | 22.9 GB | 27.4 GB | A100 | 1 | $0.462/hr |
| cheaper alt. | RTX 5060 Ti | 2 | $0.290/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Kimi-Linear-48B-A3B-Instruct at its published (BF16) precision: 5× RTX 3090, at $0.176/hr per GPU ($0.880/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Kimi-Linear-48B-A3B-Instruct: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Multi-head latent attention: every layer caches one compressed latent per token. Cached per token: 31,104 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.12 GB |
| 32K | 0.95 GB |
| 128K | 3.8 GB |
| 1M (model maximum) | 30.4 GB |
Computed from the layer, head and window counts in the model's own configuration (moonshotai/Kimi-Linear-48B-A3B-Instruct), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
Kimi-Linear-48B-A3B-Instruct: common questions
Can Kimi-Linear-48B-A3B-Instruct run on a single GPU?
No. At BF16 it needs 109.8 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 24.0 GB RTX 3090, and it takes 5 of them.
How many GPUs do I need to run Kimi-Linear-48B-A3B-Instruct?
5 at BF16. It needs 109.8 GB of VRAM and the cheapest capable live offer is a 24.0 GB RTX 3090, so 5 of them come to $0.880/hr in total.
What is the least VRAM Kimi-Linear-48B-A3B-Instruct can run in?
27.4 GB, at INT4 (quantized), which fits a 32 GB card, against 109.8 GB at BF16. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.
Does quantizing Kimi-Linear-48B-A3B-Instruct lower the GPU bill?
Yes. At BF16 the cheapest live fit is 5 RTX 3090 cards at $0.880/hr. At INT4 (quantized) it drops to one A100 at $0.462/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
Alternatives at this size
Other models for text-generation within about a third of Kimi-Linear-48B-A3B-Instruct's 49.1B parameters, from other model lines.
- Qwen-72B (72.3B, BF16)
- Llama-3.3-70B-Instruct (70.6B, BF16)
- Ornith-1.5-35B-A3B (36.0B, BF16)
- Phi-3.5-MoE-instruct (41.9B, BF16)
- Laguna-XS-2.1 (33.4B, BF16)
More on Kimi-Linear-48B-A3B-Instruct
Fits on a 141 GB GPU at BF16: every model that fits in 141 GB.
Fits on an 80 GB GPU at FP8: every model that fits in 80 GB.
Fits on a 32 GB GPU at INT4: every model that fits in 32 GB.
Best chat and assistants models: how Kimi-Linear-48B-A3B-Instruct ranks against the rest.
Related reading: RTX 3090 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
Your own GPU cloud accounts
One console for the GPU clouds you already rent from. Your keys stay on your machine.
$20 a month, flat