OpenAI on NVIDIA Blackwell

Can I run gpt-oss 20B-A3.6B on an RTX 5070?

Not at Q4_K_M — it needs 13.1 GB against 10.5 GB available. Drop to IQ3_M and it fits, at about 176 tokens per second.

Won't fit

13.1 GB of 10.5 GB · 125%
014 GB
Weights 11.9 GB
KV cache 0.4 GB
Runtime overhead 0.8 GB
Over the limit 2.6 GB

Short by 2.6 GB. You can run it with 18 of 24 layers on the RTX 5070 and the rest in system RAM, at roughly 53.6 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation53.6tok/s
Prompt processing3617tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of gpt-oss 20B-A3.6B on a RTX 5070

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 38.9 GB 40.1 GB Won't fit — 7.46
INT8 / W8A8 8.50 20.4 GB 21.6 GB Won't fit — 17.8
FP8 (E4M3) 8.00 19.5 GB 20.6 GB Won't fit — 19.7
Q6_K 6.56 16.0 GB 17.1 GB Won't fit — 27.0
Q5_K_M 5.67 13.8 GB 15.0 GB Won't fit — 38.8
Q5_K_S 5.52 13.4 GB 14.6 GB Won't fit — 39.8
AWQ 4-bit 4.25 11.9 GB 13.1 GB Won't fit — 53.5
GPTQ 4-bit 4.25 11.9 GB 13.1 GB Won't fit — 53.5
MXFP4 4.25 11.9 GB 13.1 GB Won't fit — 53.5
Q4_K_M 4.85 11.9 GB 13.1 GB Won't fit — 53.6
Q4_K_S 4.58 11.3 GB 12.4 GB Won't fit — 62.7
Q4_0 4.55 11.2 GB 12.4 GB Won't fit — 63.1
IQ4_XS 4.25 10.5 GB 11.7 GB Won't fit — 86.8
Q3_K_M 3.91 9.7 GB 10.9 GB Won't fit — 109
IQ3_M 3.70 9.2 GB 10.4 GB Fits, but tight 9K 176
IQ3_XXS 3.06 7.8 GB 8.9 GB Runs comfortably 41K 199
Q2_K 2.63 6.8 GB 8.0 GB Runs comfortably 62K 219
IQ2_XXS 2.06 5.5 GB 6.7 GB Runs comfortably 90K 251
IQ1_M 1.75 4.8 GB 5.9 GB Runs comfortably 105K 273

Also worth checking

gpt-oss 20B-A3.6B on every GPU
Full hardware tableEverything that runs on a RTX 5070
Full model tableWhy the KV cache is the part that bites
Guide