Mistral on NVIDIA Blackwell

Can I run Mixtral 8x7B on an RTX 5090?

Yes. Mixtral 8x7B at Q4_K_M uses 28.2 GB of the 29.3 GB available on a RTX 5090, and runs at about 120 tokens per second. You can push the context to 17K.

Fits, but tight

28.2 GB of 29.3 GB · 96%
029 GB
Weights 26.4 GB
KV cache 1.0 GB
Runtime overhead 0.8 GB

This fits with almost nothing to spare. A background application claiming VRAM will push it over. Drop to the next quantisation down, or quantise the KV cache to Q8_0 — that halves the cache for no meaningful quality loss.

Generation120tok/s
Prompt processing3402tok/s
Max context17Ktokens
KV per 1K tokens0GB

Every quantisation of Mixtral 8x7B on a RTX 5090

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 87.0 GB 88.8 GB Won't fit — 2.45
INT8 / W8A8 8.50 46.2 GB 48.0 GB Won't fit — 7.35
Q8_0 (GGUF) 8.50 46.2 GB 48.0 GB Won't fit — 7.35
FP8 (E4M3) 8.00 43.5 GB 45.3 GB Won't fit — 8.36
Q6_K 6.56 35.7 GB 37.5 GB Won't fit — 14.4
Q5_K_M 5.67 30.8 GB 32.6 GB Won't fit — 28.5
Q5_K_S 5.52 30.0 GB 31.8 GB Won't fit — 35.7
Q4_K_M 4.85 26.4 GB 28.2 GB Fits, but tight 17K 120
Q4_K_S 4.58 24.9 GB 26.7 GB Fits, but tight 28K 126
Q4_0 4.55 24.8 GB 26.6 GB Fits, but tight 30K 127
AWQ 4-bit 4.25 23.5 GB 25.3 GB Runs comfortably 32K 133
GPTQ 4-bit 4.25 23.5 GB 25.3 GB Runs comfortably 32K 133
MXFP4 4.25 23.5 GB 25.3 GB Runs comfortably 32K 133
IQ4_XS 4.25 23.1 GB 25.0 GB Runs comfortably 32K 134
Q3_K_M 3.91 21.3 GB 23.1 GB Runs comfortably 32K 144
IQ3_M 3.70 20.2 GB 22.0 GB Runs comfortably 32K 151
IQ3_XXS 3.06 16.7 GB 18.5 GB Runs comfortably 32K 175
Q2_K 2.63 14.4 GB 16.2 GB Runs comfortably 32K 197
IQ2_XXS 2.06 11.3 GB 13.1 GB Runs comfortably 32K 236
IQ1_M 1.75 9.6 GB 11.4 GB Runs comfortably 32K 264

Also worth checking

Mixtral 8x7B on every GPU
Full hardware tableEverything that runs on a RTX 5090
Full model tableWhy the KV cache is the part that bites
Guide