Qwen on NVIDIA Ada

Can I run Qwen3 30B-A3B on an RTX 4070 Ti Super?

Not at Q4_K_M — it needs 18.8 GB against 14.2 GB available. Drop to IQ3_XXS and it fits, at about 116 tokens per second.

Won't fit

18.8 GB of 14.2 GB · 132%
020 GB
Weights 17.3 GB
KV cache 0.8 GB
Runtime overhead 0.8 GB
Over the limit 4.6 GB

Short by 4.6 GB. You can run it with 35 of 48 layers on the RTX 4070 Ti Super and the rest in system RAM, at roughly 40.2 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation40.2tok/s
Prompt processing5533tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Qwen3 30B-A3B on a RTX 4070 Ti Super

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 56.9 GB 58.4 GB Won't fit 6.92
INT8 / W8A8 8.50 30.1 GB 31.6 GB Won't fit 15.5
Q8_0 (GGUF) 8.50 30.1 GB 31.6 GB Won't fit 15.5
FP8 (E4M3) 8.00 28.4 GB 30.0 GB Won't fit 16.7
Q6_K 6.56 23.3 GB 24.9 GB Won't fit 22.9
Q5_K_M 5.67 20.2 GB 21.7 GB Won't fit 29.4
Q5_K_S 5.52 19.6 GB 21.2 GB Won't fit 31.1
Q4_K_M 4.85 17.3 GB 18.8 GB Won't fit 40.2
Q4_K_S 4.58 16.3 GB 17.9 GB Won't fit 45.8
Q4_0 4.55 16.2 GB 17.8 GB Won't fit 46.0
AWQ 4-bit 4.25 16.0 GB 17.5 GB Won't fit 48.9
GPTQ 4-bit 4.25 16.0 GB 17.5 GB Won't fit 48.9
MXFP4 4.25 16.0 GB 17.5 GB Won't fit 48.9
IQ4_XS 4.25 15.2 GB 16.7 GB Won't fit 56.0
Q3_K_M 3.91 14.0 GB 15.5 GB Won't fit 70.4
IQ3_M 3.70 13.3 GB 14.8 GB Won't fit 2K 83.3
IQ3_XXS 3.06 11.1 GB 12.6 GB Runs comfortably 26K 116
Q2_K 2.63 9.6 GB 11.1 GB Runs comfortably 42K 122
IQ2_XXS 2.06 7.6 GB 9.1 GB Runs comfortably 63K 132
IQ1_M 1.75 6.5 GB 8.0 GB Runs comfortably 74K 139

Also worth checking