Ollama VRAM Requirements and RTX 3090 Benchmarks

Compare 601 local LLM profiles by minimum VRAM, optimal VRAM, RTX 3090 speed, system RAM target, and whether the model is safe to run locally or better rented on cloud GPUs.

Start here if you are searching for "how much VRAM does this Ollama model need", "can it run on RTX 3090", or "which local LLM fits 24GB VRAM".

Ollama naming/source alignment date: 2026-02-24. Each profile links back to its Ollama library source.

Use VRAM Calculator Compatibility Matrix Multimodal Hub Sample Compare Page

Fast Answers for High-Intent Searches

601

Total profiles

16

Group hubs

0

Measured profiles

601

Estimated profiles

238

SC keywords recycled

19

Top curated models

Top 20 Curated (Ollama Popular)

Rank Model Tag VRAM Q4 VRAM Q5 System RAM RTX 3090 fit Category Command
1 Qwen3 Coder qwen3-coder:30b 20GB 22GB 32GB+ Comfortable Coding ollama run qwen3-coder:30b
2 Qwen3.6 qwen3.6:35b 22GB 24GB 32GB+ Comfortable Reasoning ollama run qwen3.6:35b
3 Qwen3.5 qwen3.5:35b 22GB 24GB 32GB+ Comfortable Reasoning ollama run qwen3.5:35b
4 GPT-OSS gpt-oss:20b 16GB 20GB 32GB+ Comfortable Reasoning ollama run gpt-oss:20b
6 Gemma4 gemma4:26b 18GB 22GB 32GB+ Comfortable Multimodal / General ollama run gemma4:26b
7 Llama 4 llama4:16x17b 40GB 48GB 64GB+ Cloud / larger GPU Multimodal / MoE ollama run llama4:16x17b
8 Qwen3 qwen3:32b 20GB 24GB 32GB+ Comfortable Coding / Reasoning ollama run qwen3:32b
9 Qwen3 qwen3:14b 12GB 14GB 32GB+ Comfortable Coding / Reasoning ollama run qwen3:14b
10 Qwen3 qwen3:8b 8GB 10GB 16GB+ Comfortable Coding / Chat ollama run qwen3:8b
11 Magistral magistral:24b 18GB 22GB 32GB+ Comfortable Reasoning ollama run magistral:24b
12 Ministral 3 ministral-3:14b 12GB 14GB 32GB+ Comfortable Reasoning / Chat ollama run ministral-3:14b
13 Nemotron 3 Nano nemotron-3-nano:30b 20GB 24GB 32GB+ Comfortable Reasoning / Chat ollama run nemotron-3-nano:30b
14 Qwen3.5 qwen3.5:122b 70GB 80GB 128GB+ Cloud / larger GPU Reasoning / Large ollama run qwen3.5:122b
15 Qwen3 Coder qwen3-coder:480b 215GB 230GB 128GB+ Cloud / larger GPU Coding / Large ollama run qwen3-coder:480b
16 GPT-OSS gpt-oss:120b 70GB 80GB 128GB+ Cloud / larger GPU Reasoning / Large ollama run gpt-oss:120b
17 Qwen3 VL qwen3-vl:32b 20GB 24GB 32GB+ Comfortable Multimodal / Vision ollama run qwen3-vl:32b
18 Llama 3.3 llama3.3:70b 40GB 48GB 64GB+ Cloud / larger GPU General Chat ollama run llama3.3:70b
19 DeepSeek-R1 deepseek-r1:32b 20GB 24GB 32GB+ Comfortable Reasoning / Math ollama run deepseek-r1:32b
20 Gemma 3 gemma3:12b 12GB 14GB 32GB+ Comfortable Multimodal / General ollama run gemma3:12b

Group Hubs

Featured high-demand models

Model VRAM (min/optimal) System RAM RTX 3090 fit 3090 tok/s Path
Llama 4 128X17B Q4 418GB / 428GB 128GB+ Cloud / larger GPU 1.1 View
Llama 4 128X17B Q5 420GB / 430GB 128GB+ Cloud / larger GPU 1 View
Llama 4 128X17B Q8 424GB / 434GB 128GB+ Cloud / larger GPU 0.8 View
Llama 4 128X17B FP16 430GB / 442GB 128GB+ Cloud / larger GPU 0.6 View
DeepSeek-R1 671B Q4 418GB / 428GB 128GB+ Cloud / larger GPU 1.1 View
DeepSeek-R1 671B Q5 420GB / 430GB 128GB+ Cloud / larger GPU 1 View
DeepSeek-R1 671B Q8 424GB / 434GB 128GB+ Cloud / larger GPU 0.8 View
DeepSeek-R1 671B FP16 430GB / 442GB 128GB+ Cloud / larger GPU 0.6 View

Search Console feedback loop

Top keywords feeding daily content and model page updates.

Keyword Clicks Impressions CTR Landing
deepseek r1 model sizes and ram requirements 0 3 0% /ar/models/deepseek-r1-70b-q8/
deepseek 70b q8 0 1 0% /ar/models/deepseek-r1-70b-q8/
deepseek coder 6.7b 0 1 0% /ar/models/deepseek-coder-67b-q8/
deepseek-v3 vram requirements 0 1 0% /ar/models/deepseek-v3-671b-q8/
how much vram to load llama 2 70b fp16 0 1 0% /ar/models/llama2-70b-fp16/
ollama llama 3 8b memory requirements 0 1 0% /ar/models/llama3-8b-q4/
ollama qwen3 14b model size memory requirements 0 1 0% /ar/models/qwen3-14b-q4/
ai83090 0 141 0% /de/updates/
vram 0 6 0% /de/tools/vram-calculator/
cluster lokale gruppe 0 3 0% /de/guides/ollama-local-cluster-network/
conversion trichter 0 3 0% /de/status/conversion-funnel/
rtx 3090 vram 0 3 0% /de/hardware/verified-3090/
vram 0 3 0% /de/
vram testen 0 3 0% /de/tools/vram-calculator/
3090 vram 0 2 0% /de/hardware/verified-3090/
rtx 3090 tokens per second 0 2 0% /de/hardware/verified-3090/
vram benchmark 0 2 0% /de/
vram test 0 2 0% /de/
vram test 0 2 0% /de/tools/vram-calculator/
vram testen 0 2 0% /de/
vram tester 0 2 0% /de/
vram tester 0 2 0% /de/tools/vram-calculator/
vram testing 0 2 0% /de/
3090 benchmark 0 1 0% /de/hardware/verified-3090/
cluster lokale konten 0 1 0% /de/guides/ollama-local-cluster-network/
cuda out of memory 0 1 0% /de/errors/cuda-out-of-memory/
qwen3-coder 30b q4 0 1 0% /de/models/qwen3-coder-30b-q4/
qwen4b 0 1 0% /de/models/qwen-4b-q8/
ram rechner 0 1 0% /de/tools/vram-calculator/
rtx 3090 benchmark 0 1 0% /de/hardware/verified-3090/
test vram 0 1 0% /de/
test vram 0 1 0% /de/tools/vram-calculator/
testing vram 0 1 0% /de/
testing vram 0 1 0% /de/tools/vram-calculator/
vram 3090 0 1 0% /de/hardware/verified-3090/
vram testing 0 1 0% /de/tools/vram-calculator/
vramtest 0 1 0% /de/tools/vram-calculator/
ai83090 0 6033 0% /en/blog/
ai83090 0 4006 0% /en/models/
ai83090 0 2468 0% /en/status/
ai83090 0 1029 0% /en/updates/
best local llm 16gb vram 0 18 0% /en/blog/16gb-vram-model-selection-guide/
llama4:16x17b 0 7 0% /en/models/llama4-16x17b-q8/
16gb vram llm 0 4 0% /en/blog/16gb-vram-model-selection-guide/
ollama cluster 0 3 0% /en/guides/ollama-local-cluster-network/
deepseek coder v2 hardware requirements 0 2 0% /en/models/deepseek-coder-v2-16b-q5/
mistral nemo 12b vram requirements 0 2 0% /en/models/mistral-nemo-12b-q4/
my vram 0 2 0% /en/tools/vram-calculator/
qwen3 14b vram requirements 0 2 0% /en/models/qwen3-14b-q5/
qwen3-vl-8b-instruct vram requirements 0 2 0% /en/models/qwen3-vl-8b-q5/
rtx 4090 llm 0 2 0% /en/blog/rtx4090-vs-rtx3090-for-local-llm/
vram benchmark 0 2 0% /en/
vram site 0 2 0% /en/
"gpt-oss:120b-cloud" 0 1 0% /en/models/gpt-oss-120b-cloud/
%site.ollama.com library qwen3.5 4b 9b 0 1 0% /en/models/
%site.ollama.com/library qwen3-coder 30b 0 1 0% /en/models/qwen3-coder-30b-q5/
+site.ollama.com library qwen3.5 4b 9b 0 1 0% /en/models/
3090 local llm 0 1 0% /en/blog/rtx4090-vs-rtx3090-for-local-llm/
3090 ti vram 0 1 0% /en/hardware/verified-3090/
38077 0 1 0% /en/hardware/verified-3090/
apple silicon unified memory local llm vram shared memory 0 1 0% /en/hardware/apple-silicon-llm-guide/
best llm for 16gb vram 0 1 0% /en/blog/16gb-vram-model-selection-guide/
best llm for 16gb vram 2026 0 1 0% /en/blog/16gb-vram-model-selection-guide/
best local llm for 16gb vram 0 1 0% /en/blog/16gb-vram-model-selection-guide/
best local llm model for 16gb vram 0 1 0% /en/blog/16gb-vram-model-selection-guide/
best local model for 16gb vram 0 1 0% /en/blog/16gb-vram-model-selection-guide/
best models for 24gb vram 0 1 0% /en/blog/best-24gb-vram-models-2026/
best qwen for coding 0 1 0% /en/guides/best-coding-models/
dolphin3:8b 0 1 0% /en/models/dolphin3-8b-q5/
dolphin3:8b 0 1 0% /en/models/dolphin3-8b-q8/
gemma 2 9b vram requirements 0 1 0% /en/models/gemma2-9b-q8/
gemma 3 12b vram requirements 0 1 0% /en/models/gemma3-12b-q4/
gemma 3 27b vram requirements 0 1 0% /en/models/gemma3-27b-q5/
gpt oss 20b vram requirements 0 1 0% /en/models/gpt-oss-20b-fp16/
https://ollama.com/library/qwen2.5-vl 0 1 0% /en/models/qwen25vl-7b-q5/
llama 2 7b memory requirements fp16 ram vram 0 1 0% /en/models/
local rag on laptop 2gb ram 2026 0 1 0% /en/guides/best-rag-models/
mixtral 8x7b vram requirements 0 1 0% /en/models/mixtral-8x7b-q4/
ollama 16gb vram 0 1 0% /en/models/
ollama calculator 0 1 0% /en/tools/vram-calculator/
ollama deepseek-r1:8b model tag 0 1 0% /en/models/deepseek-r1-8b-q8/
ollama gpu calculator 0 1 0% /en/tools/vram-calculator/
ollama qwen2.5 14b memory requirements 0 1 0% /en/models/
ollama qwen2.5-coder 14b model size 0 1 0% /en/models/qwen25-coder-14b-q5/
ollama run deepseek-coder:6.7b 0 1 0% /en/models/deepseek-coder-67b-fp16/
ollama vs vllm 0 1 0% /en/guides/ollama-vs-vllm-vram/
q8 vs q4 0 1 0% /en/tools/quantization-blind-test/
qwen2.5-coder 14b vram requirements 0 1 0% /en/models/qwen25-coder-14b-q5/
qwen2.5-vl 7b vram requirements 0 1 0% /en/models/qwen25vl-7b-q5/
qwen2.5-vl-7b vram requirements 0 1 0% /en/models/qwen25vl-7b-q5/
qwen2.5:1.5b ollama 0 1 0% /en/models/qwen25-15b-q8/
qwen3.5-397b-a17b vram requirements 0 1 0% /en/models/qwen35-397b-a17b-cloud/
qwen3.8 q4 vs q8 0 1 0% /en/blog/en-tools-quantization-blind-test/
rtx 3090 ti vram 0 1 0% /en/hardware/verified-3090/
what is minimum requirement 0 1 0% /en/models/qwen25vl-7b-q5/
yes 0 1 0% /en/models/gemma2-27b-fp16/
どれぐらいですか 0 1 0% /en/models/qwen25vl-7b-q5/
ai83090 0 1182 0% /es/status/
llama 2 7b fp16 vram requirement 0 5 0% /es/models/llama2-7b-fp16/
rtx 3090 en 2026 0 5 0% /es/blog/still-the-vram-king-rtx-3090-2026/
vram 0 4 0% /es/
vram 0 4 0% /es/tools/vram-calculator/
rtx 3090 2026 0 3 0% /es/blog/still-the-vram-king-rtx-3090-2026/
how much vram needed for llama 2 7b fp16 0 2 0% /es/models/llama2-7b-fp16/
llama 2 7b memory requirements fp16 vram 0 2 0% /es/models/llama2-7b-fp16/
rtx 3090 vram 0 2 0% /es/blog/still-the-vram-king-rtx-3090-2026/
clusters locales 0 1 0% /es/guides/ollama-local-cluster-network/
como saber mi vram 0 1 0% /es/
como se cuanta vram tengo 0 1 0% /es/tools/vram-calculator/
cuantificado 0 1 0% /es/tools/quantization-blind-test/
how much vram ram needed to run llama 2 7b fp16 0 1 0% /es/models/llama2-7b-fp16/
llama4:16x17b 0 1 0% /es/hardware/verified-3090/
llama4:16x17b 0 1 0% /es/models/
llama4:16x17b 0 1 0% /es/status/conversion-funnel/
llama4:16x17b 0 1 0% /es/tools/vram-calculator/
llama4:16x17b 0 1 0% /es/updates/
nvidia 3090 vram 0 1 0% /es/blog/still-the-vram-king-rtx-3090-2026/
ollama deepseek-r1 8b model tag official 0 1 0% /es/models/deepseek-r1-8b-q8/
ollama qwen2.5-coder 14b model size 0 1 0% /es/models/qwen25-coder-14b-fp16/
quantifying 0 1 0% /es/tools/quantization-blind-test/
rag list 0 1 0% /es/guides/best-rag-models/
rtx 3090 vram 0 1 0% /es/hardware/verified-3090/
test vram 0 1 0% /es/
test vram 0 1 0% /es/tools/vram-calculator/
v ram 0 1 0% /es/
v ram 0 1 0% /es/tools/vram-calculator/
vllm vs ollama 0 1 0% /es/guides/ollama-vs-vllm-vram/
vram test 0 1 0% /es/
vram test 0 1 0% /es/tools/vram-calculator/
vrams 0 1 0% /es/
ai83090 0 379 0% /fr/updates/
deepseek r1 model sizes and ram requirements 0 13 0% /fr/models/deepseek-r1-8b-q5/
cuda out of memory 0 6 0% /fr/errors/cuda-out-of-memory/
apple silicon llm 0 4 0% /fr/hardware/apple-silicon-llm-guide/
ollama vs vllm 0 4 0% /fr/guides/ollama-vs-vllm-vram/
location gpu 0 3 0% /fr/affiliate/cloud-gpu/
local llm benchmark 0 2 0% /fr/blog/
cuda out of memory error 0 1 0% /fr/errors/cuda-out-of-memory/
location carte graphique 0 1 0% /fr/affiliate/cloud-gpu/
ninfer 3090 0 1 0% /fr/hardware/verified-3090/
ollama qwen3 coder 30b model size 0 1 0% /fr/models/qwen3-coder-30b-fp16/
ollama qwen3-coder 30b model size 0 1 0% /fr/models/qwen3-coder-30b-fp16/
ollama qwen3-coder 30b size 0 1 0% /fr/models/qwen3-coder-30b-fp16/
vram 0 1 0% /fr/
vram 0 1 0% /fr/tools/vram-calculator/
vram estimator 0 1 0% /fr/tools/vram-calculator/
rtx 3090 24gb ram 0 1 0% /hi/models/llava-7b-q4/
ai83090 0 538 0% /id/updates/
deepseek r1 model sizes and ram requirements 0 4 0% /id/models/deepseek-r1-15b-q4/
ollama llama 3.1 8b ram requirements 0 3 0% /id/models/llama31-8b-q4/
vram 0 2 0% /id/tools/vram-calculator/
ollama llama 3 8b memory requirements 0 1 0% /id/models/llama3-8b-q8/
ollama qwen2.5-coder 14b model size 0 1 0% /id/models/qwen25-coder-14b-q5/
ollama qwen2.5-coder 14b size 0 1 0% /id/models/qwen25-coder-14b-q5/
ollama qwen2.5:1.5b size 0 1 0% /id/models/qwen25-15b-q8/
qwen3 1.7b q4 0 1 0% /id/models/qwen3-17b-q4/
qwen3 8b system requirements 0 1 0% /id/models/qwen3-8b-q5/
ai83090 0 129 0% /ja/blog/
ai83090 0 81 0% /ja/status/
データ 鮮度 0 26 0% /ja/status/data-freshness/
データ鮮度 0 13 0% /ja/status/data-freshness/
cuda out of memory 0 4 0% /ja/errors/cuda-out-of-memory/
ブラインドテスト 0 3 0% /ja/tools/quantization-blind-test/
コンバージョンファネルとは 0 2 0% /ja/status/conversion-funnel/
cuda error out of memory 0 1 0% /ja/errors/cuda-out-of-memory/
gpt-oss-120b vram requirements 0 1 0% /ja/models/gpt-oss-120b-q5/
llama 3.1 405b vram requirements 0 1 0% /ja/models/llama31-405b-q5/
ollama qwen2.5:1.5b size 0 1 0% /ja/models/qwen25-15b-q5/
コンバージョンファネル 0 1 0% /ja/status/conversion-funnel/
ai83090 0 47 0% /ko/blog/
데이터 신선도 0 42 0% /ko/status/data-freshness/
70b q4 0 4 0% /ko/models/codellama-70b-q4/
3090 vram 0 3 0% /ko/hardware/verified-3090/
dolphin3:8b 0 3 0% /ko/models/dolphin3-8b-q4/
qwen2.5-coder:1.5b 0 2 0% /ko/models/qwen25-coder-15b-q5/
qwen3 14b q4 0 2 0% /ko/models/qwen3-14b-q4/
qwen3-14b q4 0 2 0% /ko/models/qwen3-14b-q4/
블라인드 테스트 0 2 0% /ko/tools/quantization-blind-test/
local llm vram calculator 1 1 100% /ko/tools/vram-calculator/
3090 24gb vram 0 1 0% /ko/hardware/verified-3090/
32b 0 1 0% /ko/models/qwen25-coder-32b-q4/
7b q4 0 1 0% /ko/models/mistral-7b-q4/
cuda error out of memory 0 1 0% /ko/errors/cuda-out-of-memory/
deepseek-coder-v2 0 1 0% /ko/models/deepseek-coder-v2-16b-q4/
dolphin 3 8b 0 1 0% /ko/models/dolphin3-8b-q4/
dolphin3 8b 0 1 0% /ko/models/dolphin3-8b-q4/
gemma4 e2b q4 0 1 0% /ko/models/gemma4-e2b-q4/
gpt-oss-20b q4 0 1 0% /ko/models/gpt-oss-20b-q4/
granite 3.1 moe 0 1 0% /ko/models/granite31-moe-3b-q4/
llama4:16x17b 0 1 0% /ko/models/llama4-16x17b-fp16/
mistral 7b q4 0 1 0% /ko/models/mistral-7b-q4/
ollama 16gb vram 0 1 0% /ko/hardware/16gb-vram-models/
qwen2 5 coder 0 1 0% /ko/models/qwen25-coder-14b-fp16/
qwen2 5 coder 0 1 0% /ko/models/qwen25-coder-15b-q5/
qwen2 5 coder 0 1 0% /ko/models/qwen25-coder-7b-q5/
qwen2.5-coder:1.5b 0 1 0% /ko/models/qwen25-coder-15b-fp16/
qwen3 4b 0 1 0% /ko/models/qwen3-4b-q4/
qwen3 4b 0 1 0% /ko/models/qwen3-4b-q8/
qwen3 4b q4 0 1 0% /ko/models/qwen3-4b-q4/
qwen3-4b 0 1 0% /ko/models/qwen3-4b-q4/
qwen3-4b 0 1 0% /ko/models/qwen3-4b-q8/
qwen3-vl-2b 0 1 0% /ko/models/qwen3-vl-2b-cloud/
rtx 3090 vram 0 1 0% /ko/hardware/verified-3090/
rtx3090 vram 0 1 0% /ko/hardware/verified-3090/
vram 3090 0 1 0% /ko/hardware/verified-3090/
vram calculator 0 1 0% /ko/tools/vram-calculator/
vram-calculator 0 1 0% /ko/tools/vram-calculator/
ai83090 0 355 0% /pt/blog/
ai83090 0 105 0% /pt/status/
vram 0 3 0% /pt/tools/vram-calculator/
vram 0 2 0% /pt/
localllama 0 1 0% /pt/blog/
ollama qwen2.5 7b model size official 0 1 0% /pt/models/qwen25-7b-fp16/
sim 0 1 0% /pt/tools/vram-calculator/
v ram 0 1 0% /pt/
v ram 0 1 0% /pt/tools/vram-calculator/
ai83090 0 781 0% /ru/updates/
cuda error out of memory при майнинге 0 36 0% /ru/errors/cuda-out-of-memory/
cuda ошибка out of memory 0 24 0% /ru/blog/cuda-out-of-memory-ollama-fixes-2026/
cuda ошибка out of memory 0 24 0% /ru/errors/cuda-out-of-memory/
qwen3 4b vram requirements 0 3 0% /ru/models/qwen3-4b-q5/
vram 0 3 0% /ru/tools/vram-calculator/
deepseek r1 model sizes and ram requirements 0 2 0% /ru/models/deepseek-r1-8b-q8/
3090 2026 0 1 0% /ru/blog/still-the-vram-king-rtx-3090-2026/
dolphin 3.0 r1 mistral 24b 0 1 0% /ru/tools/vram-calculator/
gpt oss 120b vram requirements 0 1 0% /ru/models/gpt-oss-120b-cloud/
gpt-oss-120b vram requirements 0 1 0% /ru/models/gpt-oss-120b-cloud/
mixtral 8x7b vram requirements 0 1 0% /ru/models/mixtral-8x7b-q8/
ollama deepseek-r1:8b model tag 0 1 0% /ru/models/deepseek-r1-8b-q8/
ollama qwen3:8b size 0 1 0% /ru/models/qwen3-8b-q5/
q4 vs q8 0 1 0% /ru/blog/en-tools-quantization-blind-test/
qwen3 235b vram requirements 0 1 0% /ru/models/qwen3-235b-q8/
qwen3 32b vram requirements 0 1 0% /ru/models/qwen3-32b-q8/
qwen3 4b requirements 0 1 0% /ru/models/qwen3-4b-q5/
qwen3 8b vram requirements 0 1 0% /ru/models/qwen3-8b-q5/
qwen3:8b requirements 0 1 0% /ru/models/qwen3-8b-q5/
rtx 3090 vram 0 1 0% /ru/blog/still-the-vram-king-rtx-3090-2026/
vram калькулятор 0 1 0% /ru/