GeForce RTX 4090

Name: GeForce RTX 4090
Brand: NVIDIA
Price: 4395 USD

NVIDIA GeForce RTX 4090: 24 GB of memory at 1008 GB/s. Below: the open models it can run, estimated at 24 GB with an 8K context. 24 GB of VRAM at 1,008 GB/s: it runs 27–32B dense models at 4-bit, and MoE models with about 3B active parameters generate quickly.

Last updated: 2026-10-14

Specifications

Memory	24 GB GDDR6X[src]
Memory bandwidth	1008 GB/s[src]
Power	450 W[src]
Launch	2022-10[src]
Launch price	US$1,599[src]
Architecture	Ada Lovelace[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)	Q4_K_M	21.4 GiB	388
Qwen3.6-35B-A3B	35B (3B)	Q4_K_M	21.4 GiB	388
Qwen3-32B	32.8B	Q4_K_M	21.9 GiB	35
DeepSeek-R1-Distill-Qwen-32B	32.76B	Q4_K_M	21.9 GiB	36
Gemma 4 31B IT	30.7B	Q5_K_M	22.9 GiB	32
Qwen3-30B-A3B (2507)	30.5B (3.3B)	Q5_K_M	22.5 GiB	301
Qwen3-Coder-30B-A3B-Instruct	30.5B (3.3B)	Q5_K_M	22.5 GiB	301
GLM-4.7-Flash (30B-A3B)	30B (3B)	Q5_K_M	21.8 GiB	331
Qwen3.5-27B	27B	Q6_K	22.7 GiB	32
Qwen3.6-27B	27B	Q6_K	22.7 GiB	32
Qwen3.8-27B	27B	Q6_K	22.7 GiB	32
Gemma 3 27B IT	27B	Q6_K	23.3 GiB	32
Gemma 4 26B A4B IT	25.2B (3.8B)	Q6_K	21.1 GiB	225
Mistral Small 3.2 24B Instruct (2506)	24B	Q6_K	21.1 GiB	36
Magistral Small 1.2 (2509)	24B	Q6_K	21.1 GiB	36
Devstral Small 2 24B Instruct (2512)	24B	Q6_K	21.1 GiB	36
gpt-oss-20b	21B (3.6B)	Q8_0	22.5 GiB	184
Phi-4 (14B)	14B	Q8_0	16.6 GiB	47
Gemma 3 12B IT	12B	Q8_0	13.8 GiB	55
Gemma 4 12B Unified IT	11.95B	Q8_0	13.3 GiB	56
Qwen3.5-9B	9B	FP16	18.4 GiB	39
Qwen3-8B	8.2B	FP16	17.7 GiB	43
DeepSeek-R1-0528-Qwen3-8B	8.19B	FP16	17.6 GiB	43
Llama 3.1 8B Instruct	8B	FP16	17.1 GiB	44
Gemma 4 E4B IT	8B	FP16	16.4 GiB	44
Llama 3.2 3B Instruct	3.21B	FP16	7.7 GiB	110
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	NeweggAlliance Group (USA)	MSI Gaming (MSI) GeForce RTX 4090 24GB GDDR6X PCI Express 4.0 Video Card RTX 4090 GAMING X TRIO 24GNewegg marketplace seller	US$4,395	US$183/GB	2026-10-14

US value: about 145.5 tok/s on Llama 3.1 8B Instruct (Q4_K_M, 8K), or 3.3 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 4090 run?

With 24 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 4090 for local LLMs?

Generation speed is mostly set by memory bandwidth (1008 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 44 tokens/s; bigger models are slower.

Can the GeForce RTX 4090 run 70B models?

Llama 3.3 70B Instruct does not fit on the GeForce RTX 4090 (24 GB), even at Q4_K_M with an 8K context.