Qwen on NVIDIA Ada

Can I run Qwen3 235B-A22B on an L40S?

Not at Q4_K_M — it needs 135.1 GB against 44.3 GB available. You would need 4 of these cards.

Won't fit

135.1 GB of 44.3 GB · 305%
0141 GB
Weights 132.8 GB
KV cache 1.5 GB
Runtime overhead 0.8 GB
Over the limit 90.8 GB

Short by 90.8 GB. You can run it with 29 of 94 layers on the L40S and the rest in system RAM, at roughly 3.90 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation3.90tok/s
Prompt processing1717tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Qwen3 235B-A22B on a L40S

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 437.9 GB 440.2 GB Won't fit 0.99
INT8 / W8A8 8.50 232.3 GB 234.6 GB Won't fit 2.00
Q8_0 (GGUF) 8.50 232.3 GB 234.6 GB Won't fit 2.00
FP8 (E4M3) 8.00 218.9 GB 221.2 GB Won't fit 2.14
Q6_K 6.56 179.5 GB 181.8 GB Won't fit 2.70
Q5_K_M 5.67 155.2 GB 157.5 GB Won't fit 3.21
Q5_K_S 5.52 151.1 GB 153.4 GB Won't fit 3.33
Q4_K_M 4.85 132.8 GB 135.1 GB Won't fit 3.90
Q4_K_S 4.58 125.5 GB 127.8 GB Won't fit 4.22
Q4_0 4.55 124.7 GB 127.0 GB Won't fit 4.24
AWQ 4-bit 4.25 118.0 GB 120.3 GB Won't fit 4.58
GPTQ 4-bit 4.25 118.0 GB 120.3 GB Won't fit 4.58
MXFP4 4.25 118.0 GB 120.3 GB Won't fit 4.58
IQ4_XS 4.25 116.5 GB 118.8 GB Won't fit 4.64
Q3_K_M 3.91 107.2 GB 109.5 GB Won't fit 5.21
IQ3_M 3.70 101.5 GB 103.8 GB Won't fit 5.63
IQ3_XXS 3.06 84.1 GB 86.4 GB Won't fit 7.52
Q2_K 2.63 72.4 GB 74.7 GB Won't fit 9.83
IQ2_XXS 2.06 56.9 GB 59.2 GB Won't fit 16.6
IQ1_M 1.75 48.4 GB 50.7 GB Won't fit 26.8

Also worth checking