Qwen on NVIDIA Ampere

Can I run Qwen3 14B on an RTX 3080 10GB?

Not at Q4_K_M — it needs 10.5 GB against 8.6 GB available. Drop to IQ3_XXS and it fits, at about 73.8 tokens per second.

Won't fit

10.5 GB of 8.6 GB · 123%
011 GB
Weights 8.5 GB
KV cache 1.3 GB
Runtime overhead 0.8 GB
Over the limit 1.9 GB

Short by 1.9 GB. You can run it with 30 of 40 layers on the RTX 3080 10GB and the rest in system RAM, at roughly 14.0 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation14.0tok/s
Prompt processing853tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Qwen3 14B on a RTX 3080 10GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 27.5 GB 29.6 GB Won't fit 1.79
INT8 / W8A8 8.50 14.3 GB 16.4 GB Won't fit 4.56
Q8_0 (GGUF) 8.50 14.3 GB 16.4 GB Won't fit 4.56
FP8 (E4M3) 8.00 13.8 GB 15.8 GB Won't fit 4.72
Q6_K 6.56 11.3 GB 13.4 GB Won't fit 7.08
Q5_K_M 5.67 9.7 GB 11.8 GB Won't fit 9.50
Q5_K_S 5.52 9.5 GB 11.6 GB Won't fit 10.3
AWQ 4-bit 4.25 9.4 GB 11.5 GB Won't fit 10.4
GPTQ 4-bit 4.25 9.4 GB 11.5 GB Won't fit 10.4
MXFP4 4.25 9.4 GB 11.5 GB Won't fit 10.4
Q4_K_M 4.85 8.5 GB 10.5 GB Won't fit 14.0
Q4_K_S 4.58 8.0 GB 10.1 GB Won't fit 17.2
Q4_0 4.55 8.0 GB 10.1 GB Won't fit 17.3
IQ4_XS 4.25 7.5 GB 9.6 GB Won't fit 1K 22.0
Q3_K_M 3.91 7.0 GB 9.1 GB Won't fit 5K 34.0
IQ3_M 3.70 6.7 GB 8.8 GB Won't fit 7K 41.7
IQ3_XXS 3.06 5.7 GB 7.8 GB Fits, but tight 13K 73.8
Q2_K 2.63 5.0 GB 7.1 GB Runs comfortably 17K 82.3
IQ2_XXS 2.06 4.2 GB 6.3 GB Runs comfortably 23K 97.0
IQ1_M 1.75 3.7 GB 5.8 GB Runs comfortably 26K 107

Also worth checking