GeForce RTX 5090

Name: GeForce RTX 5090
Brand: NVIDIA
Price: 6885.9 USD

NVIDIA GeForce RTX 5090: 32 GB of memory at 1792 GB/s. Below: the open models it can run, estimated at 32 GB with an 8K context. 32 GB of VRAM at 1,792 GB/s, the most of any GeForce card here: 32B models fit at 6-bit, or at 4-bit with room for longer contexts.

Last updated: 2026-10-14

Specifications

Memory	32 GB GDDR7[src]
Memory bandwidth	1792 GB/s[src]
Power	575 W[src]
Launch	2025-01[src]
Launch price	US$1,999[src]
Architecture	Blackwell[src]
Software	CUDA

What it can run

Best-quality quantization that fits with a 8K context, and the estimated generation speed. Estimates, not benchmarks. See the methodology.

Model	Params	Best quant	Memory	tok/s (est.)
Qwen3.5-35B-A3B	35B (3B)	Q6_K	28.9 GiB	508
Qwen3.6-35B-A3B	35B (3B)	Q6_K	28.9 GiB	508
Qwen3-32B	32.8B	Q6_K	28.9 GiB	46
DeepSeek-R1-Distill-Qwen-32B	32.76B	Q6_K	28.9 GiB	46
Gemma 4 31B IT	30.7B	Q6_K	26.3 GiB	50
Qwen3-30B-A3B (2507)	30.5B (3.3B)	Q6_K	25.8 GiB	461
Qwen3-Coder-30B-A3B-Instruct	30.5B (3.3B)	Q6_K	25.8 GiB	461
GLM-4.7-Flash (30B-A3B)	30B (3B)	Q6_K	25.1 GiB	508
Qwen3.5-27B	27B	Q8_0	29.1 GiB	44
Qwen3.6-27B	27B	Q8_0	29.1 GiB	44
Qwen3.8-27B	27B	Q8_0	29.1 GiB	44
Gemma 3 27B IT	27B	Q8_0	29.6 GiB	44
Gemma 4 26B A4B IT	25.2B (3.8B)	Q8_0	27.0 GiB	311
Mistral Small 3.2 24B Instruct (2506)	24B	Q8_0	26.7 GiB	49
Magistral Small 1.2 (2509)	24B	Q8_0	26.7 GiB	49
Devstral Small 2 24B Instruct (2512)	24B	Q8_0	26.7 GiB	49
gpt-oss-20b	21B (3.6B)	Q8_0	22.5 GiB	328
Phi-4 (14B)	14B	FP16	29.4 GiB	45
Gemma 3 12B IT	12B	FP16	24.8 GiB	52
Gemma 4 12B Unified IT	11.95B	FP16	24.2 GiB	52
Qwen3.5-9B	9B	FP16	18.4 GiB	70
Qwen3-8B	8.2B	FP16	17.7 GiB	76
DeepSeek-R1-0528-Qwen3-8B	8.19B	FP16	17.6 GiB	77
Llama 3.1 8B Instruct	8B	FP16	17.1 GiB	78
Gemma 4 E4B IT	8B	FP16	16.4 GiB	78
Llama 3.2 3B Instruct	3.21B	FP16	7.7 GiB	195
Kimi K2.6	1000B (32B)	Doesn't fit	–	–
DeepSeek-R1-0528 (671B)	671B (37B)	Doesn't fit	–	–
DeepSeek-V4-Flash	284B (13B)	Doesn't fit	–	–
Qwen3.8-Flash-Next	180B (6B)	Doesn't fit	–	–
Qwen3.5-122B-A10B	122B (10B)	Doesn't fit	–	–
Mistral Small 4 119B (2603)	119B (6.5B)	Doesn't fit	–	–
gpt-oss-120b	117B (5.1B)	Doesn't fit	–	–
Llama 4 Scout 17B-16E Instruct	109B (17B)	Doesn't fit	–	–
GLM-4.5-Air	106B (12B)	Doesn't fit	–	–
Qwen3-Coder-Next (80B-A3B)	80B (3B)	Doesn't fit	–	–
Llama 3.3 70B Instruct	70B	Doesn't fit	–	–

Prices

Latest price at each shop, as listed on the day we checked.

Region	Shop	Listing	Price	Per GB	Checked
US	NeweggSlava computers	Gigabyte GeForce RTX 5090 WINDFORCE OC 32G GV-N5090WF3OC-32GD GDDR7 PCIe 5.0 x16 video cardNewegg marketplace seller; ships from Israel	US$6,885.90	US$215/GB	2026-10-14

US value: about 258.6 tok/s on Llama 3.1 8B Instruct (Q4_K_M, 8K), or 3.8 tok/s per US$100.

A price history chart appears once we have checked this device on two or more days.

All prices →

Frequently asked questions

Which LLMs can the GeForce RTX 5090 run?

With 32 GB and an 8K context, the largest models that fit include: Qwen3.5-35B-A3B, Qwen3.6-35B-A3B, Qwen3-32B, DeepSeek-R1-Distill-Qwen-32B, Gemma 4 31B IT. See the full table below.

How fast is the GeForce RTX 5090 for local LLMs?

Generation speed is mostly set by memory bandwidth (1792 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 78 tokens/s; bigger models are slower.

Can the GeForce RTX 5090 run 70B models?

Llama 3.3 70B Instruct does not fit on the GeForce RTX 5090 (32 GB), even at Q4_K_M with an 8K context.