Skip to content

DeepSeek V2 Lite on 2× RTX 3090: VRAM, speed and fit

DeepSeek V2 Lite at 8-bit runs on the 2× RTX 3090 with 160K of context at about 168 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on 2× RTX 3090 · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
21.6 GB
of 48.0 GB
Context
160K
auto
Speed
168 tok/s
73 when full
First reply
0.3 s
2K prompt

Where the memory goes

Model
15.5 GB
Context
4.7 GB
Overhead
1.3 GB
OS
0.0 GB
Free
26.4 GB
Memory
21.6 GB of 48.0 GB
Context
160K tokens · auto
Speed
168 tok/s · 73 when full
First reply
0.3 s for a 2K prompt
Running cost
$11/mo at home · $33/mo on a cloud RTX A5000

Validation: Validated

What DeepSeek V2 Lite on 2× RTX 3090 was checked against
CheckTargetResultStatus
Fits in GPU memory48.0 GB usable21.6 GB neededPass
Headroom5% free26.4 GB free (55%)Pass
Context for chat16K or more160KPass
Writing speed15 tok/s or more168 tok/sPass
First replyunder 3 s, 2K prompt0.3 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
DeepSeek V2 Lite
Parameters
15.7B, 2.7B active
Precision
8-bit (Q8_0)
Context
160K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
2× RTX 3090
GPU can use
48.0 GB of 48.0 GB

Running cost

At home
$11/mo
500 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 legraphista/DeepSeek-V2-Lite-IMat-GGUF
vLLM
vllm serve deepseek-ai/DeepSeek-V2-Lite

DeepSeek V2 Lite on similar machines

All machines: DeepSeek V2 Lite VRAM requirements. Everything that runs here: what can the 2× RTX 3090 run?

Other models that run on 2× RTX 3090

Questions

Can DeepSeek V2 Lite run on 2× RTX 3090?

DeepSeek V2 Lite at 8-bit runs on the 2× RTX 3090 with 160K of context at about 168 tokens a second.

Report VR-0BPC9ES, calculated Nov 22, 2026 from published specs, not benchmarked.