Phi-3 Mini 3.8B Instruct on NVIDIA RTX PRO 6000 Blackwell

Yes — RTX PRO 6000 Blackwell runs Phi-3 Mini 3.8B Instruct excellently at Q8_0 — 257 tok/s. 96 GB VRAM with plenty of headroom.

Model Size

3.82B

Device VRAM

96 GB

Bandwidth

1800 GB/s

Quantization

Q8_0

Benchmarks

Performance by Quantization

OwnRig currently has one published compatibility entry for Phi-3 Mini 3.8B Instruct on NVIDIA RTX PRO 6000 Blackwell at Q8_0. This is the best supported pairing we can stand behind today.

Quantization	Speed	TTFT	Fits in VRAM	Rating	Confidence
Q8_0	257 tok/s	15ms	✓ Yes	Excellent	estimated

Notes

Q8_0

96 GB GDDR7 fits Q8_0 (4.5 GB) with 92 GB headroom. 1.8 TB/s bandwidth. Near-FP16 quality — minimal quantization loss.

About Phi-3 Mini 3.8B Instruct

Phi-3 Mini 3.8B Instruct (3.82B) is a chat, coding, reasoning model. Punches above its weight: a 3.8B model that rivals many 7B models on reasoning benchmarks. MIT license. Well suited for resource-constrained setups or as a fast secondary model.

View all Phi-3 Mini 3.8B Instruct hardware options →

About NVIDIA RTX PRO 6000 Blackwell

NVIDIA RTX PRO 6000 Blackwell has 96 GB at 1800 GB/s. Street price: $7,500.

See all models NVIDIA RTX PRO 6000 Blackwell can run →

Estimate method: Estimated from bandwidth ratio vs RTX 4090 (1800/1008 = 1.78x) with conservative margin. Reference hardware source: nvidia.com (2026-03-29)

Performance varies by driver version, inference engine, quantization method, context length, and system configuration. Figures shown are estimates based on community benchmarks and may not reflect your exact setup. Product names are trademarks of their respective owners. OwnRig is independent and not affiliated with any hardware or AI model provider.