Can Kimi Linear 48B A3B run on NVIDIA H20 96GB?

YES — Runs Great

A84Great

Estimated from fit model

Kimi Linear 48B A3B needs ~42.2 GB VRAM. NVIDIA H20 96GB has 96.0 GB. With Q4_K_M quantization, expect ~89 tok/s.

Runtime: vLLMCapacity: RoomyBandwidth: HighStack: OptimizedBottleneck: Balanced

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 42.2 GB, 88.5 tok/s, Runs well

42.2 GB required96.0 GB available

44% VRAM used

Fit status

Runs well

Decode

88.5 tok/s

TTFT

2187 ms

Safe context

944K

Memory

42.2 GB / 96.0 GB

Memory breakdown

Weights29.3 GB

KV Cache0.9 GB

Runtime2.4 GB

Headroom9.6 GB

See how fast it feels

See how fast it feelsKimi Linear 48B A3B on NVIDIA H20 96GB
1st promptCold start — includes initialization
>Write a Python function to sort a list
Here's a Python function that sorts a list using quicksort:

def quicksort(arr):
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left   = [x for x in arr if x < pivot]
    middle = [x for x in arr if x == pivot]
    right  = [x for x in arr if x > pivot]
    return quicksort(left) + middle + quicksort(right)

# Example usage
numbers = [3, 6, 8, 10, 1, 2, 1]
print(quicksort(numbers))  # [1, 1, 2, 3, 6, 8, 10]

The algorithm picks a pivot, partitions the list into smaller, equal, and larger elements, then recursively sorts each partition and concatenates the results.
2nd promptWarm — model ready, no init overhead
>Now add type hints and a docstring
def quicksort(arr: list[int]) -> list[int]:
    """Sort a list using the quicksort algorithm.

    Args:
        arr: List of integers to sort.

    Returns:
        A new sorted list.
    """
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left   = [x for x in arr if x < pivot]
    middle = [x for x in arr if x == pivot]
    right  = [x for x in arr if x > pivot]
    return quicksort(left) + middle + quicksort(right)

Added type hints for the input and return type, plus a Google-style docstring.
Estimated: 88.5 tok/s decode · 2.2s TTFT (warm) · 221 tok/s prefill

What limits this setup

This setup is broadly balanced for this model.

No major red flags

This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.

Best improvement path

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	A	Runs well	88.5 tok/s	1193 ms	944K
Coding	A	Runs well	88.5 tok/s	2187 ms	944K
Agentic Coding	A	Runs well	88.5 tok/s	3181 ms	944K
Reasoning	A	Runs well	88.5 tok/s	2585 ms	944K
RAG	A	Runs well	88.5 tok/s	3976 ms	944K

Inference speed

Kimi Linear 48B A3B inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Kimi Linear 48B A3B at Q4_K_M across popular GPUs and Apple Silicon, including multi-GPU rigs, using the fastest local runtime per device. Fastest is 2× RX 7900 XTX 24GB at ~40 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.

GPU / Mac	Memory	Quant	Speed (tok/s)	Fits?
2× RX 7900 XTX 24GB	48 GB	Q4_K_M	40.1	Fits
2× RTX 4090 24GB	48 GB	Q4_K_M	29.3	Fits
RTX 5090 32GB	32 GB	Q4_K_M	25.8	Heavy offload
2× RTX 3090 24GB	48 GB	Q4_K_M	25.1	Fits
4× RTX 3060 12GB	48 GB	Q4_K_M	22.1	Fits
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	21.1	Fits
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	21.1	Tight
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	19.0	Fits
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	15.8	Fits
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	15.0	Fits
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	10.6	Offloads
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	8.2	Tight
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	7.5	Tight
RTX 4090 24GB	24 GB	Q4_K_M	7.1	Too big
RX 7900 XTX 24GB	24 GB	Q4_K_M	6.4	Too big
RTX 3090 24GB	24 GB	Q4_K_M	6.1	Too big
RTX 4080 Super 16GB	16 GB	Q4_K_M	2.5	Too big
RTX 4070 12GB	12 GB	Q4_K_M	2.0	Too big
RTX 3060 12GB	12 GB	Q4_K_M	2.0	Too big
RTX 4060 8GB	8 GB	Q4_K_M	2.0	Too big

Estimates for single-stream decoding at Q4_K_M; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.

Quantization options

How Kimi Linear 48B A3B (48B params) fits at each quantization level on NVIDIA H20 96GB (96.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	18.7 GB	Low	A73
Q3_K_S	3	23.5 GB	Low	A74
NVFP4	4	26.9 GB	Medium	A74
Q4_K_M	4	29.3 GB	Medium	A75
Q5_K_M	5	34.6 GB	High	A76
Q6_K	6	39.4 GB	High	A77
Q8_0Best for your GPU	8	51.4 GB	Very High	A80
F16	16	98.4 GB	Maximum	F0

Get started

Copy-paste commands to run Kimi Linear 48B A3B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "moonshotai/Kimi-Linear-48B-A3B-Instruct" \
  --hf-file "Kimi-Linear-48B-A3B-Instruct-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Your hardware

More models your NVIDIA H20 96GB can run

Model	Params	Grade	Decode
Qwen 3.5 122B A10B	122B	S	71.3 tok/s
Mistral Small 4 119B	119B	S	77.3 tok/s
GPT-OSS 120B	117B	S	39.5 tok/s
Command A 111B	111B	S	41.8 tok/s
Qwen 2.5 VL 72B	72B	S	64.2 tok/s

Frequently asked questions

Can NVIDIA H20 96GB run Kimi Linear 48B A3B?

Yes, NVIDIA H20 96GB can run Kimi Linear 48B A3B with a A grade (Runs well). Expected decode speed: 88.5 tok/s.

How much VRAM does Kimi Linear 48B A3B need?

Kimi Linear 48B A3B (48B parameters) requires approximately 42.2 GB of memory with Q4_K_M quantization.

What is the best quantization for Kimi Linear 48B A3B?

The recommended quantization for Kimi Linear 48B A3B is Q4_K_M, which balances quality and memory efficiency.

What speed will Kimi Linear 48B A3B run at on NVIDIA H20 96GB?

On NVIDIA H20 96GB, Kimi Linear 48B A3B achieves approximately 88.5 tokens per second decode speed with a time-to-first-token of 2187ms using Q4_K_M quantization.

Can NVIDIA H20 96GB run Kimi Linear 48B A3B for coding?

For coding workloads, Kimi Linear 48B A3B on NVIDIA H20 96GB receives a A grade with 88.5 tok/s and 944K context.

What context window can Kimi Linear 48B A3B use on NVIDIA H20 96GB?

On NVIDIA H20 96GB, Kimi Linear 48B A3B can safely use up to 944K tokens of context. The model's official context limit is 1.0M, but available memory constrains the safe maximum.

See all results for NVIDIA H20 96GB See all hardware for Kimi Linear 48B A3B

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/kimi-linear-48b-a3b-on-h20-96gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview:

Workload

Grade

Fit

Decode

TTFT

Context

Chat

Runs well

88.5 tok/s

1193 ms

944K

Coding

Runs well

88.5 tok/s

2187 ms

944K

Agentic Coding

Runs well

88.5 tok/s

3181 ms

944K

Reasoning

Runs well

88.5 tok/s

2585 ms

944K

RAG

Runs well

88.5 tok/s

3976 ms

944K

Kimi Linear 48B A3B inference speed — tokens per second by GPU & Mac

GPU / Mac

Memory

Quant

Speed (tok/s)

Fits?

2× RX 7900 XTX 24GB

48 GB

Q4_K_M

40.1

Fits

2× RTX 4090 24GB

48 GB

Q4_K_M

29.3

Fits

RTX 5090 32GB

32 GB

Q4_K_M

25.8

Heavy offload

2× RTX 3090 24GB

48 GB

Q4_K_M

25.1

Fits

4× RTX 3060 12GB

48 GB

Q4_K_M

22.1

Fits

MacBook Pro M4 Max 128GB

128 GB

Q4_K_M

21.1

Fits

MacBook Pro M4 Max 64GB

64 GB

Q4_K_M

21.1

Tight

Mac Studio M3 Ultra 256GB

256 GB

Q4_K_M

19.0

Fits

Mac Studio M2 Ultra 128GB

128 GB

Q4_K_M

15.8

Fits

Mac Studio M1 Ultra 128GB

128 GB

Q4_K_M

15.0

Fits

MacBook Pro M4 Pro 48GB

48 GB

Q4_K_M

10.6

Offloads

MacBook Pro M3 Max 64GB

64 GB

Q4_K_M

8.2

Tight

MacBook Pro M1 Max 64GB

64 GB

Q4_K_M

7.5

Tight

RTX 4090 24GB

24 GB

Q4_K_M

7.1

Too big

RX 7900 XTX 24GB

24 GB

Q4_K_M

6.4

Too big

RTX 3090 24GB

24 GB

Q4_K_M

6.1

Too big

RTX 4080 Super 16GB

16 GB

Q4_K_M

2.5

Too big

RTX 4070 12GB

12 GB

Q4_K_M

2.0

Too big

RTX 3060 12GB

12 GB

Q4_K_M

2.0

Too big

RTX 4060 8GB

8 GB

Q4_K_M

2.0

Too big

Quantization options

How Kimi Linear 48B A3B (48B params) fits at each quantization level on NVIDIA H20 96GB (96.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	18.7 GB	Low	A73
Q3_K_S	3	23.5 GB	Low	A74
NVFP4	4	26.9 GB	Medium	A74
Q4_K_M	4	29.3 GB	Medium	A75
Q5_K_M	5	34.6 GB	High	A76
Q6_K	6	39.4 GB	High	A77
Q8_0Best for your GPU	8	51.4 GB	Very High	A80
F16	16	98.4 GB	Maximum	F0

Model

Params

Grade

Decode

Capabilities

Qwen 3.5 122B A10B

122B

71.3 tok/s

Mistral Small 4 119B

119B

77.3 tok/s

GPT-OSS 120B

117B

39.5 tok/s

Command A 111B

111B

41.8 tok/s

Qwen 2.5 VL 72B

72B

64.2 tok/s

Frequently asked questions

Can NVIDIA H20 96GB run Kimi Linear 48B A3B?

Yes, NVIDIA H20 96GB can run Kimi Linear 48B A3B with a A grade (Runs well). Expected decode speed: 88.5 tok/s.

How much VRAM does Kimi Linear 48B A3B need?

Kimi Linear 48B A3B (48B parameters) requires approximately 42.2 GB of memory with Q4_K_M quantization.

What is the best quantization for Kimi Linear 48B A3B?

The recommended quantization for Kimi Linear 48B A3B is Q4_K_M, which balances quality and memory efficiency.

What speed will Kimi Linear 48B A3B run at on NVIDIA H20 96GB?

On NVIDIA H20 96GB, Kimi Linear 48B A3B achieves approximately 88.5 tokens per second decode speed with a time-to-first-token of 2187ms using Q4_K_M quantization.

Can NVIDIA H20 96GB run Kimi Linear 48B A3B for coding?

For coding workloads, Kimi Linear 48B A3B on NVIDIA H20 96GB receives a A grade with 88.5 tok/s and 944K context.

What context window can Kimi Linear 48B A3B use on NVIDIA H20 96GB?

On NVIDIA H20 96GB, Kimi Linear 48B A3B can safely use up to 944K tokens of context. The model's official context limit is 1.0M, but available memory constrains the safe maximum.