How much VRAM does Mixtral_34Bx2_MoE_60B need?

Q4_K_M is exactly 34,390,405,728 bytes (32.03 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Mixtral_34Bx2_MoE_60B's KV cache?

7.50 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

Is Mixtral_34Bx2_MoE_60B a mixture-of-experts model?

Yes — 2 experts with 2 routed per token. Every expert must be resident in memory, but only the routed ones are read per token, so its memory requirement and its speed behave very differently.

cloudyu · text · mixture of experts

Mixtral_34Bx2_MoE_60B

cloudyu/Mixtral_34Bx2_MoE_60B

Mixtral_34Bx2_MoE_60B at Q4_K_M is exactly 34,390,405,728 bytes (32.03 GiB / 34.39 GB) — an effective 4.524 bits per weight, not the nominal 4. Its KV cache at 32K is 7.50 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

60.8B

total, not active

Architecture

llama

60 layers

Context

200,000

native (config.json)

License

cc-by-nc-4.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	19.14 GiB	20,546,130,528	2.703	—	TheBloke
Q3_K_M	24.95 GiB	26,787,709,536	3.524	—	TheBloke
Q4_0	31.98 GiB	34,333,635,168	4.517	—	TheBloke
Q4_K_M	32.03 GiB	34,390,405,728	4.524	—	TheBloke
Q5_0	39.00 GiB	41,877,811,808	5.509	—	TheBloke
Q5_K_M	39.03 GiB	41,907,057,248	5.513	—	TheBloke
Q6_K	46.47 GiB	49,893,499,488	6.563	—	TheBloke

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.94 GiB	0.94 GiB	—	60 / 0 / 0
8,192	1.88 GiB	1.88 GiB	—	60 / 0 / 0
16,384	3.75 GiB	3.75 GiB	—	60 / 0 / 0
32,768	7.50 GiB	7.50 GiB	—	60 / 0 / 0
65,536	15.00 GiB	15.00 GiB	—	60 / 0 / 0
131,072	30.00 GiB	30.00 GiB	—	60 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs Qwen3.6-27B vs Qwen3.8-27B vs gemma-4-E4B-it vs Qwen3-30B-A3B-Thinking-2507 vs Qwen3-8B All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 31.86 GiB. The real file is 32.03 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

128

Hidden size

7168

Vocab

64,000

Sliding window

none

SWA period

—

MLA

Experts

Experts per token

use_sliding_window

—

Questions people ask

How much VRAM does Mixtral_34Bx2_MoE_60B need?: Q4_K_M is exactly 34,390,405,728 bytes (32.03 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Mixtral_34Bx2_MoE_60B's KV cache?: 7.50 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Is Mixtral_34Bx2_MoE_60B a mixture-of-experts model?: Yes — 2 experts, 2 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of Mixtral_34Bx2_MoE_60B should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.