What is the biggest model the RTX 5090 can run?

Qwen3 32B, which needs about 22 GB, on the 32 GB version at about 49 tok/s.

What is the biggest model the RTX Pro 6000 Blackwell can run?

gpt-oss 120B, which needs about 67 GB, on the 96 GB version at about 220 tok/s.

RTX 5090 vs RTX Pro 6000 Blackwell for local AI

The RTX Pro 6000 Blackwell fits more: up to gpt-oss 120B, against Qwen3 32B on the RTX 5090. They generate at about the same speed, since their memory bandwidth is close (1792 and 1792 GB/s).

The numbers that matter

Memory and bandwidth

Memory decides which models fit. Bandwidth decides how fast they generate, because every token reads the model's active weights once.

	RTX 5090	RTX Pro 6000 Blackwell
Memory	32 GB VRAM	96 GB VRAM
Bandwidth	1792 GB/s	1792 GB/s
Biggest model that fits	Qwen3 32B	gpt-oss 120B
Best everyday pick	Qwen3 32B, ~49 tok/s	gpt-oss 120B, ~220 tok/s

Model by model

Every model, side by side

Estimated generation speed at the largest memory size each was sold in (32 GB and 96 GB).

Model	RTX 5090	RTX Pro 6000 Blackwell
Qwen3 4B About 4 GB	~390 tok/s Runs great	~390 tok/s Runs great
Qwen3 8B About 7 GB	~190 tok/s Runs great	~190 tok/s Runs great
Gemma 3 12B About 10 GB	~120 tok/s Runs great	~120 tok/s Runs great
Qwen3 14B About 11 GB	~110 tok/s Runs great	~110 tok/s Runs great
gpt-oss 20B About 16 GB	~310 tok/s Runs great	~310 tok/s Runs great
Gemma 3 27B About 19 GB	~58 tok/s Runs great	~58 tok/s Runs great
Qwen3 30B-A3B About 21 GB	~310 tok/s Runs great	~310 tok/s Runs great
Qwen3 32B About 22 GB	~49 tok/s Runs great	~49 tok/s Runs great
Llama 3.3 70B About 45 GB	Needs 45 GB	~23 tok/s Usable
gpt-oss 120B About 67 GB	Needs 67 GB	~220 tok/s Runs great
Qwen3 235B-A22B About 144 GB	Needs 144 GB	Needs 144 GB
DeepSeek R1 671B About 406 GB	Needs 406 GB	Needs 406 GB

Which should you get?

It depends. The RTX Pro 6000 Blackwell fits bigger models; the RTX 5090 runs the models both fit faster. Price matters as much as either number: check used prices on the builds page, where every parts list has a measured or estimated speed.

How we estimate

Both columns use the same formula: memory bandwidth divided by the weights read per token, calibrated on our own rack, with Ollama's default 4-bit sizes. Real results are usually within about a third. Measured either one? Send us the number.

Compare

Other matchups

Every comparison