what/llm/can/i/run
models/qwen-qwen3-30b-a3b-instruct

Qwen3 30B A3B 2507

Sources arena · Confidence C
params
30B
dense
bench sources
1
single
context
max tokens
license
Apache 2.0
permissive

Model meta

canonical
Qwen/Qwen3-30B-A3B-Instruct
parameters
30B
organization
Alibaba
license
Apache 2.0
context
downloads

Benchmark breakdown

Hardware fit matrix

by gpu tierQ2_K
11.3 GB
Q4_K_M
15.0 GB
Q5_K_M
18.8 GB
Q6_K
22.5 GB
Q8_0
30.0 GB
FP16
60.0 GB
G8 GB tier
rtx 3050 · 4060
G12 GB tier
rtx 3060 · 4070
G16 GB tier
rtx 4080 · 4060 ti 16g
GREAT
49 t/s
G24 GB tier
rtx 3090 · 4090
GREAT
79 t/s
GREAT
60 t/s
GREAT
48 t/s
G32 GB tier
rtx 5090 · m3 max
GREAT
146 t/s
GREAT
111 t/s
GREAT
89 t/s
GREAT
75 t/s
G48 GB tier
a6000 · m3 max 64
GREAT
47 t/s
GREAT
36 t/s
GREAT
29 t/s
GREAT
24 t/s
GREAT
18 t/s
G80 GB tier
h100 · m3 ultra 128
GREAT
160 t/s
GREAT
121 t/s
GREAT
98 t/s
GREAT
82 t/s
GREAT
62 t/s
GREAT
31 t/s
full fit · production speed tight fit · usabledoesn't fit
Get personalized ranking
Tell us your machine — we'll tell you if this is actually your best pick, or what's better.
Rank for my hardware