What can NVIDIA RTX 2080 Ti 22GB (China-mod) run for creative?

Build: NVIDIA RTX 2080 Ti 22GB (China-mod) + — + 32 GB RAM (windows)

Memory: 22 GB VRAM + 32 GB system RAM
Runner: llama.cpp / Ollama (CUDA)

Runs comfortably
94 models

Ranked by fit for creative use case + predicted speed. Click a row for VRAM breakdown.

#1Hermes 3 Llama 3.1 8B
8B
hermes
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 14.7 GBHeadroom: 7.3 GB
ollama run hermes3:8b
47
tok/s
Estimated
Weights
8.50 GB
KV cache
4.00 GB
Activations
0.43 GB
Runtime
1.80 GB
#2Gemma 2 9B Instruct
9B
gemma
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 16.6 GBHeadroom: 5.4 GB
ollama run gemma2:9b
42
tok/s
Estimated
Weights
9.80 GB
KV cache
4.50 GB
Activations
0.50 GB
Runtime
1.80 GB
#3Dolphin 3.0 Llama 3.2 3B
3B
dolphin
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 5.2 GBHeadroom: 16.8 GB
221
tok/s
Estimated
Weights
1.80 GB
KV cache
1.50 GB
Activations
0.10 GB
Runtime
1.80 GB
#4Hermes 3 Llama 3.2 3B
3B
hermes
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 5.2 GBHeadroom: 16.8 GB
221
tok/s
Estimated
Weights
1.80 GB
KV cache
1.50 GB
Activations
0.10 GB
Runtime
1.80 GB
#5Gemma 2 2B Instruct
2.61B
gemma
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 4.9 GBHeadroom: 17.1 GB
254
tok/s
Estimated
Weights
1.70 GB
KV cache
1.30 GB
Activations
0.09 GB
Runtime
1.80 GB
#6Gemma 3 4B
4B
gemma
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 8.4 GBHeadroom: 13.6 GB
ollama run gemma3:4b
94
tok/s
Estimated
Weights
4.40 GB
KV cache
2.00 GB
Activations
0.23 GB
Runtime
1.80 GB
#7Qwen 3 1.7B
1.7B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 3.8 GBHeadroom: 18.2 GB
390
tok/s
Estimated
Weights
1.10 GB
KV cache
0.85 GB
Activations
0.06 GB
Runtime
1.80 GB
#8Llama 3.2 3B Instruct
3B
llama
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 6.9 GBHeadroom: 15.1 GB
ollama run llama3.2:3b
126
tok/s
Estimated
Weights
3.40 GB
KV cache
1.50 GB
Activations
0.18 GB
Runtime
1.80 GB
#9Qwen2-VL 2B Instruct
2B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 4.0 GBHeadroom: 18.0 GB
332
tok/s
Estimated
Weights
1.10 GB
KV cache
1.00 GB
Activations
0.06 GB
Runtime
1.80 GB
#10Kumru 2B
2.4B
mistral
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 4.6 GBHeadroom: 17.4 GB
ollama run alibayram/kumru:latest
276
tok/s
Estimated
Weights
1.50 GB
KV cache
1.20 GB
Activations
0.08 GB
Runtime
1.80 GB
#11Phi-3.5 Vision
4.2B
phi
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 6.5 GBHeadroom: 15.5 GB
158
tok/s
Estimated
Weights
2.50 GB
KV cache
2.10 GB
Activations
0.13 GB
Runtime
1.80 GB
#12Qwen 3.5 2B Turkish SFT
2B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 4.2 GBHeadroom: 17.8 GB
332
tok/s
Estimated
Weights
1.30 GB
KV cache
1.00 GB
Activations
0.07 GB
Runtime
1.80 GB

Runs with tradeoffs
31 models

Tight VRAM, partial CPU offload, or context-limited.

Qwen 3 30B-A3B
30B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 35.7 GBHeadroom: 5.5 GB
  • • Partial CPU offload: ~38% of layers run on CPU
ollama run qwen3:30b
201
tok/s
Estimated
Weights
18.00 GB
KV cache
15.00 GB
Activations
0.91 GB
Runtime
1.80 GB
DeepSeek Coder V2 Lite (16B)
16B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 19.8 GBHeadroom: 2.2 GB
  • • Tight VRAM fit — only 2.2 GB headroom left for context growth
ollama run deepseek-coder-v2:16b
276
tok/s
Estimated
Weights
9.50 GB
KV cache
8.00 GB
Activations
0.48 GB
Runtime
1.80 GB
DeepSeek V2 Lite Chat
15.7B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 18.7 GBHeadroom: 3.3 GB
  • • Tight VRAM fit — only 3.3 GB headroom left for context growth
276
tok/s
Estimated
Weights
8.60 GB
KV cache
7.85 GB
Activations
0.44 GB
Runtime
1.80 GB
Dolphin 3.0 Mistral 24B
24B
dolphin
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 19.5 GBHeadroom: 2.5 GB
  • • Tight VRAM fit — only 2.5 GB headroom left for context growth
28
tok/s
Estimated
Weights
14.00 GB
KV cache
3.00 GB
Activations
0.70 GB
Runtime
1.80 GB
Jamba 1.5 Mini
52B
other
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 40.9 GBHeadroom: 0.3 GB
  • • Partial CPU offload: ~46% of layers run on CPU
55
tok/s
Estimated
Weights
31.00 GB
KV cache
6.50 GB
Activations
1.55 GB
Runtime
1.80 GB
Mixtral 8x7B Instruct
47B
mixtral
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 37.1 GBHeadroom: 4.1 GB
  • • Partial CPU offload: ~41% of layers run on CPU
ollama run mixtral:8x7b
51
tok/s
Estimated
Weights
28.00 GB
KV cache
5.88 GB
Activations
1.40 GB
Runtime
1.80 GB
Qwen 2.5 14B Instruct
14B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 18.2 GBHeadroom: 3.8 GB
  • • Tight VRAM fit — only 3.8 GB headroom left for context growth
ollama run qwen2.5:14b
47
tok/s
Estimated
Weights
8.90 GB
KV cache
7.00 GB
Activations
0.45 GB
Runtime
1.80 GB
Phi-4 Reasoning 14B
14B
phi
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 20.5 GBHeadroom: 1.5 GB
  • • Tight VRAM fit — only 1.5 GB headroom left for context growth
ollama run phi4-reasoning:14b
47
tok/s
Estimated
Weights
11.10 GB
KV cache
7.00 GB
Activations
0.56 GB
Runtime
1.80 GB

What if you upgraded?

Hypothetical scenarios. We re-ran the compatibility engine for each.

+32 GB system RAM

Check the current price

Adds 32 GB to your CPU-offload working set. Helps when models don't quite fit in VRAM.

Unlocks: 15 new comfortable, 43 new tradeoff

  • • Qwen 3 0.6B
  • • Gemma 3 270M
  • • SigLIP SO400M (patch14-384)
  • • TinyLlama 1.1B Chat v1.0

Upgrade to NVIDIA GeForce RTX 5090 Mobile

Check the current price

24 GB VRAM (vs your 22 GB) plus a bandwidth jump from ~616 GB/s to ~896 GB/s.

Unlocks: 23 new comfortable

  • • Qwen 3 0.6B
  • • Mistral Small 3 24B
  • • Gemma 3 270M
  • • Qwen 2.5 14B Instruct

Add a second NVIDIA RTX 2080 Ti 22GB (China-mod)

Check the current price

Tensor parallelism splits the model across both cards, effectively doubling VRAM. Bandwidth doesn't double — runs ~1.5× the single-card speed in practice.

Unlocks: 42 new comfortable

  • • Qwen 3 0.6B
  • • Qwen 3 30B-A3B
  • • Qwen 2.5 Coder 32B Instruct
  • • Gemma 4 31B Dense

Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.

Won't run
top 5 popular models

Need more memory than you have. Shown for orientation.

DeepSeek V4 Pro (1.6T MoE)
1600B
deepseek
Commercial OK

Even with CPU offload, needs more memory than your VRAM (22 GB) + 60% of system RAM (19 GB) combined.

—
Qwen 3 235B-A22B
235B
qwen
Commercial OK

Even with CPU offload, needs more memory than your VRAM (22 GB) + 60% of system RAM (19 GB) combined.

—
DeepSeek R1 (671B reasoning)
671B
deepseek
Commercial OK

Even with CPU offload, needs more memory than your VRAM (22 GB) + 60% of system RAM (19 GB) combined.

—
Llama 4 Scout
109B
llama
Commercial OK

Even with CPU offload, needs more memory than your VRAM (22 GB) + 60% of system RAM (19 GB) combined.

—
DeepSeek V4 Flash (284B MoE)
284B
deepseek
Commercial OK

Even with CPU offload, needs more memory than your VRAM (22 GB) + 60% of system RAM (19 GB) combined.

—
Cloud GPU rental

Not enough VRAM for these models? Rent a GPU by the hour

The models above need more memory than 22 GB VRAM + 32 GB system RAM. You can rent a GPU with enough VRAM by the hour to run one of them, or to test it before you buy hardware. For daily use, owning a card usually costs less over time.

RunPod

billed per GPU-hour

Hourly GPU pods, in community-cloud and secure-cloud tiers.

The two tiers run on different hosts and terms. Check which tier a pod is on before a long job.

Rent on RunPod

Vast.ai

billed per GPU-hour

Marketplace of host-listed GPUs; price and reliability vary by host.

Host quality varies. Filter for verified hosts and check the host's reliability before a long run.

Rent on Vast.ai

Referral links: RunLocalAI may earn a commission or account credits if you sign up and rent, at no extra cost to you. Hourly rates and which GPUs are listed change all the time, so check the provider's page before you start. Editorial opinions are independent of the referral relationship; see how we make money.

How to read these numbers

Measured here
Measured here - RunLocalAI ran this exact combo on owner hardware with public evidence.

Source-backed
Source-backed / community - a reproduced public source supports the speed, but it is not labeled as owner-measured.

Extrapolated
Extrapolated - predicted from a measured benchmark on similar-bandwidth hardware.

Estimated
Estimated - formula based on VRAM bandwidth and model architecture; not a benchmark row.

RunLocalAI Will-It-Run Framework →

Want a specific benchmark we don't have? Email Contact support and we'll prioritize it.