What can NVIDIA GeForce RTX 5090 run for reasoning?
Build: NVIDIA GeForce RTX 5090 + — + 32 GB RAM (windows)
Memory: 32 GB VRAM + 32 GB system RAM
Runner: llama.cpp / Ollama (CUDA)
Runs comfortably152 models
Ranked by fit for reasoning use case + predicted speed. Click a row for VRAM breakdown.
Quant: Q8_0Context: 8,192VRAM: 13.8 GBHeadroom: 18.2 GBTTFT: fastollama run deepseek-r1:7b157tok/sEstimated
Quant: Q8_0Context: 8,192VRAM: 13.8 GBHeadroom: 18.2 GBTTFT: fast
ollama run deepseek-r1:7b157
tok/s
Estimated
Weights
8.10 GB
KV cache
3.50 GB
Activations
0.41 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~143 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 17.6 GBHeadroom: 14.4 GBTTFT: fastollama run phi4-reasoning:14b138tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 17.6 GBHeadroom: 14.4 GBTTFT: fast
ollama run phi4-reasoning:14b138
tok/s
Estimated
Weights
8.40 GB
KV cache
7.00 GB
Activations
0.43 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~287 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 17.6 GBHeadroom: 14.4 GBTTFT: fastollama run deepseek-r1:14b138tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 17.6 GBHeadroom: 14.4 GBTTFT: fast
ollama run deepseek-r1:14b138
tok/s
Estimated
Weights
8.40 GB
KV cache
7.00 GB
Activations
0.43 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~287 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 11.0 GBHeadroom: 21.0 GBTTFT: fast241tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 11.0 GBHeadroom: 21.0 GBTTFT: fast
241
tok/s
Estimated
Weights
4.90 GB
KV cache
4.00 GB
Activations
0.25 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~164 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 11.6 GBHeadroom: 20.4 GBTTFT: fast214tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 11.6 GBHeadroom: 20.4 GBTTFT: fast
214
tok/s
Estimated
Weights
5.00 GB
KV cache
4.50 GB
Activations
0.26 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~184 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 11.0 GBHeadroom: 21.0 GBTTFT: fastollama run RefinedNeuro/RN_TR_R1:latest241tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 11.0 GBHeadroom: 21.0 GBTTFT: fast
ollama run RefinedNeuro/RN_TR_R1:latest241
tok/s
Estimated
Weights
4.90 GB
KV cache
4.00 GB
Activations
0.25 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~164 ms (fast)
Quant: Q4_K_MContext: 8,192VRAM: 10.7 GBHeadroom: 21.3 GBTTFT: fast241tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 10.7 GBHeadroom: 21.3 GBTTFT: fast
241
tok/s
Estimated
Weights
4.70 GB
KV cache
4.00 GB
Activations
0.24 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~164 ms (fast)
Quant: Q4_K_MContext: 2,048VRAM: 23.6 GBHeadroom: 8.4 GBTTFT: noticeable62tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 23.6 GBHeadroom: 8.4 GBTTFT: noticeable
62
tok/s
Estimated
Weights
17.10 GB
KV cache
3.88 GB
Activations
0.86 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~635 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 25.8 GBHeadroom: 6.2 GBTTFT: noticeableollama run deepseek-r1:32b60tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 25.8 GBHeadroom: 6.2 GBTTFT: noticeable
ollama run deepseek-r1:32b60
tok/s
Estimated
Weights
19.00 GB
KV cache
4.00 GB
Activations
0.95 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~655 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 25.8 GBHeadroom: 6.2 GBTTFT: noticeableollama run qwq:32b60tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 25.8 GBHeadroom: 6.2 GBTTFT: noticeable
ollama run qwq:32b60
tok/s
Estimated
Weights
19.00 GB
KV cache
4.00 GB
Activations
0.95 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~655 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 24.3 GBHeadroom: 7.7 GBTTFT: noticeable60tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 24.3 GBHeadroom: 7.7 GBTTFT: noticeable
60
tok/s
Estimated
Weights
17.60 GB
KV cache
4.00 GB
Activations
0.88 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~655 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 24.3 GBHeadroom: 7.7 GBTTFT: noticeable60tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 24.3 GBHeadroom: 7.7 GBTTFT: noticeable
60
tok/s
Estimated
Weights
17.60 GB
KV cache
4.00 GB
Activations
0.88 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~655 ms (noticeable)
Runs with tradeoffs26 models
Tight VRAM, partial CPU offload, or context-limited.
Quant: Q4_K_MContext: 8,192VRAM: 28.5 GBHeadroom: 3.5 GBTTFT: fast- • Tight VRAM fit — only 3.5 GB headroom left for context growth
80tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 28.5 GBHeadroom: 3.5 GBTTFT: fast
- • Tight VRAM fit — only 3.5 GB headroom left for context growth
80
tok/s
Estimated
Weights
14.00 GB
KV cache
12.00 GB
Activations
0.71 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~492 ms (fast)
Quant: Q4_K_MContext: 2,048VRAM: 51.0 GBHeadroom: 0.2 GBTTFT: noticeable- • Partial CPU offload: ~37% of layers run on CPU
28tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 51.0 GBHeadroom: 0.2 GBTTFT: noticeable
- • Partial CPU offload: ~37% of layers run on CPU
28
tok/s
Estimated
Weights
38.50 GB
KV cache
8.75 GB
Activations
1.93 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~1434 ms (noticeable)
Quant: Q4_K_MContext: 8,192VRAM: 47.7 GBHeadroom: 3.5 GBTTFT: noticeable- • Partial CPU offload: ~33% of layers run on CPU
ollama run nemotron3:33b58tok/sEstimated
Quant: Q4_K_MContext: 8,192VRAM: 47.7 GBHeadroom: 3.5 GBTTFT: noticeable
- • Partial CPU offload: ~33% of layers run on CPU
ollama run nemotron3:33b58
tok/s
Estimated
Weights
28.00 GB
KV cache
16.50 GB
Activations
1.41 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~676 ms (noticeable)
Quant: AWQ-INT4Context: 2,048VRAM: 37.3 GBHeadroom: 13.9 GBTTFT: noticeable- • Partial CPU offload: ~14% of layers run on CPU
24tok/sEstimated
Quant: AWQ-INT4Context: 2,048VRAM: 37.3 GBHeadroom: 13.9 GBTTFT: noticeable
- • Partial CPU offload: ~14% of layers run on CPU
24
tok/s
Estimated
Weights
28.00 GB
KV cache
6.13 GB
Activations
1.40 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~1004 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 39.8 GBHeadroom: 11.4 GBTTFT: fast- • Partial CPU offload: ~20% of layers run on CPU
161tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 39.8 GBHeadroom: 11.4 GBTTFT: fast
- • Partial CPU offload: ~20% of layers run on CPU
161
tok/s
Estimated
Weights
30.00 GB
KV cache
6.50 GB
Activations
1.50 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~246 ms (fast)
Quant: Q4_K_MContext: 4,096VRAM: 28.3 GBHeadroom: 3.7 GBTTFT: noticeable- • Tight VRAM fit — only 3.7 GB headroom left for context growth
60tok/sEstimated
Quant: Q4_K_MContext: 4,096VRAM: 28.3 GBHeadroom: 3.7 GBTTFT: noticeable
- • Tight VRAM fit — only 3.7 GB headroom left for context growth
60
tok/s
Estimated
Weights
17.60 GB
KV cache
8.00 GB
Activations
0.88 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~655 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 31.4 GBHeadroom: 0.6 GBTTFT: noticeable- • Tight VRAM fit — only 0.6 GB headroom left for context growth
ollama run qwen3.6:35b-a3b55tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 31.4 GBHeadroom: 0.6 GBTTFT: noticeable
- • Tight VRAM fit — only 0.6 GB headroom left for context growth
ollama run qwen3.6:35b-a3b55
tok/s
Estimated
Weights
24.00 GB
KV cache
4.38 GB
Activations
1.20 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~717 ms (noticeable)
Quant: Q4_K_MContext: 2,048VRAM: 31.4 GBHeadroom: 0.6 GBTTFT: noticeable- • Tight VRAM fit — only 0.6 GB headroom left for context growth
ollama run qwen3.5:35b-a3b55tok/sEstimated
Quant: Q4_K_MContext: 2,048VRAM: 31.4 GBHeadroom: 0.6 GBTTFT: noticeable
- • Tight VRAM fit — only 0.6 GB headroom left for context growth
ollama run qwen3.5:35b-a3b55
tok/s
Estimated
Weights
24.00 GB
KV cache
4.38 GB
Activations
1.20 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~717 ms (noticeable)
What if you upgraded?
Hypothetical scenarios. We re-ran the compatibility engine for each.
+32 GB system RAM
~$80–150
Doubles your CPU-offload working set. Helps when models don't quite fit in VRAM.
Unlocks: 64 new comfortable, 43 new tradeoff
- • Qwen 3 0.6B
- • Llama 3.2 3B Instruct
- • Qwen 3 1.7B
- • Gemma 3 270M
Upgrade to NVIDIA A100 40GB
see current pricing
40 GB VRAM (vs your 32 GB) plus a bandwidth jump from ~1792 GB/s to ~1555 GB/s.
Unlocks: 85 new comfortable
- • Qwen 3 0.6B
- • Qwen3.6 35B-A3B
- • Llama 3.2 3B Instruct
- • Qwen 3 1.7B
Add a second NVIDIA GeForce RTX 5090
~$2499
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: 103 new comfortable
- • Qwen 3 0.6B
- • Llama 3.3 70B Instruct
- • DeepSeek R1 Distill Llama 70B
- • Qwen3.6 35B-A3B
Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.
Won't runtop 5 popular models
Need more memory than you have. Shown for orientation.
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
Even with CPU offload, needs more memory than your VRAM (32 GB) + 60% of system RAM (19 GB) combined.
—
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.Want a specific benchmark we don't have? Email Contact support and we'll prioritize it.