Skip to content

GPT Neox 20B on RTX 3090: VRAM, speed and fit

GPT Neox 20B at 8-bit runs on the RTX 3090 with 2K of context at about 29 tokens a second.

Text generation · Chat · Ollama / LM Studio

Tight fit · on RTX 3090 · 8-bit (Q8_0) · Ollama / LM Studio · Chat

97% used

GPU memory
23.3 GB
of 24.0 GB
Context
2K
auto
Speed
29 tok/s
Fast
First reply
1.3 s
2K prompt

Where the memory goes

Model
20.5 GB
Context
2.1 GB
Overhead
751 MB
OS
0.0 GB
Free
698 MB
Memory
23.3 GB of 24.0 GB
Context
2K tokens · auto
Speed
29 tok/s · fast
First reply
1.3 s for a 2K prompt
Running cost
$6.35/mo at home · $33/mo on a cloud RTX A5000

Validation: Validated with conditions

What GPT Neox 20B on RTX 3090 was checked against
CheckTargetResultStatus
Fits in GPU memory24.0 GB usable23.3 GB neededPass
Headroom5% free698 MB free (3%)Check
Context for chat2K or more2KPass
Writing speed15 tok/s or more29 tok/sPass
First replyunder 3 s, 2K prompt1.3 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
GPT Neox 20B
Parameters
20.7B
Precision
8-bit (Q8_0)
Context
2K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
RTX 3090
GPU can use
24.0 GB of 24.0 GB

Worth trying

  • RTX 5090: Runs it with 2K context
  • M5 Max · 48 GB: Runs it with 2K context
  • Ryzen AI Max+ 395 · 64 GB: Runs it with 2K context

Running cost

At home
$6.35/mo
290 W at the wall
Renting instead
$33/mo
RTX A5000 at $0.27/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 21, 2026.

How to run it

llama.cpp
llama-server -hf mradermacher/gpt-neox-20b-i1-GGUF
vLLM
vllm serve EleutherAI/gpt-neox-20b

GPT Neox 20B on similar machines

All machines: GPT Neox 20B VRAM requirements. Everything that runs here: what can the RTX 3090 run?

Other models that run on RTX 3090

Questions

Can GPT Neox 20B run on RTX 3090?

GPT Neox 20B at 8-bit runs on the RTX 3090 with 2K of context at about 29 tokens a second.

Report VR-0H71JDQ, calculated Nov 22, 2026 from published specs, not benchmarked.