RUNLOCALAIv38
→WILL IT RUNBEST GPUCOMPARETROUBLESHOOTSTARTPULSEMODELSHARDWARETOOLSBENCH
RUNLOCALAI

Independently operated catalog for local-AI hardware and software. Hand-written verdicts. Source-cited claims. Reproducible commands when we have them.

OP·Fredoline Eruo
DIR
  • Models
  • Hardware
  • Tools
  • Benchmarks
TOOLS
  • Will it run?
  • Compare hardware
  • Cost vs cloud
  • Choose my GPU
  • Quick answers
REF
  • All buyer guides
  • Methodology
  • Glossary
  • Errors KB
  • Trust
EDITOR
  • About
  • Author
  • How we make money
  • Editorial policy
  • Contact
LEGAL
  • Privacy
  • Terms
  • Sitemap
MAIL · MONTHLY DIGEST
Get monthly local AI changes
Monthly recap. No spam.
DISCLOSURE

Some links on this site are affiliate links (Amazon Associates and other first-class retailers). When you buy through them, we earn a small commission at no extra cost to you. Affiliate links do not influence our verdicts — there are cards we rate highly that we don't have affiliate relationships with, and cards that sell well that we refuse to recommend. Read more →

© 2026 runlocalai.coIndependently operated
RUNLOCALAI · v38
Will it run? / NVIDIA RTX 4090 48GB (China-mod) / reasoning

What can NVIDIA RTX 4090 48GB (China-mod) run for reasoning?

Build: NVIDIA RTX 4090 48GB (China-mod) + — + 32 GB RAM (windows)

Memory: 48 GB VRAM + 32 GB system RAM
Runner: llama.cpp / Ollama (CUDA)
AnyChatCodingAgentsReasoningVisionLong contextCreative

Runs comfortably
88 models

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

#1Llama 3.1 Nemotron Nano 8B
8B
llama
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 19.1 GBHeadroom: 28.9 GBTTFT: fast
136
tok/s
E
Weights
4.83 GB
KV cache
4.00 GB
Activations
8.43 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~248 ms (fast)
Model details →Run-on benchmark page →
#2DeepSeek R1 Distill Llama 8B
8B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 19.1 GBHeadroom: 28.9 GBTTFT: fast
136
tok/s
E
Weights
4.83 GB
KV cache
4.00 GB
Activations
8.43 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~248 ms (fast)
Model details →Run-on benchmark page →
#3DeepSeek R1 Distill Qwen 7B
7B
deepseek
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 21.3 GBHeadroom: 26.7 GBTTFT: fast
ollama run deepseek-r1:7b
88
tok/s
E
Weights
7.44 GB
KV cache
3.50 GB
Activations
8.56 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~217 ms (fast)
Model details →Run-on benchmark page →
#4Phi-4 Reasoning 14B
14B
phi
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 25.9 GBHeadroom: 22.1 GBTTFT: fast
ollama run phi4-reasoning:14b
78
tok/s
E
Weights
8.45 GB
KV cache
7.00 GB
Activations
8.61 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~434 ms (fast)
Model details →Run-on benchmark page →
#5DeepSeek R1 Distill Qwen 14B
14B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 25.9 GBHeadroom: 22.1 GBTTFT: fast
ollama run deepseek-r1:14b
78
tok/s
E
Weights
8.45 GB
KV cache
7.00 GB
Activations
8.61 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~434 ms (fast)
Model details →Run-on benchmark page →
#6DeepSeek V3 Lite (16B MoE)
16B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 28.1 GBHeadroom: 19.9 GBTTFT: instant
452
tok/s
E
Weights
9.66 GB
KV cache
8.00 GB
Activations
8.68 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~74 ms (instant)
Model details →Run-on benchmark page →
#7DeepSeek R1 Distill Mistral 24B
24B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 37.2 GBHeadroom: 10.8 GBTTFT: noticeable
45
tok/s
E
Weights
14.49 GB
KV cache
12.00 GB
Activations
8.92 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~744 ms (noticeable)
Model details →Run-on benchmark page →
#8Qwen 2.5 Math 7B
7B
qwen
Commercial OK
Quant: Q4_K_MContext: 4,096VRAM: 12.1 GBHeadroom: 35.9 GBTTFT: fast
155
tok/s
E
Weights
4.23 GB
KV cache
1.75 GB
Activations
4.31 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~217 ms (fast)
Model details →Run-on benchmark page →
#9Qwen 3 7B
7B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 17.9 GBHeadroom: 30.1 GBTTFT: fast
155
tok/s
E
Weights
4.23 GB
KV cache
3.50 GB
Activations
8.40 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~217 ms (fast)
Model details →Run-on benchmark page →
#10InternLM 2.5 7B Chat
7B
internlm
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 17.9 GBHeadroom: 30.1 GBTTFT: fast
155
tok/s
E
Weights
4.23 GB
KV cache
3.50 GB
Activations
8.40 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~217 ms (fast)
Model details →Run-on benchmark page →
#11DeepSeek R1 Distill Qwen 3 32B
32B
deepseek
Commercial OK
Quant: AWQ-INT4Context: 2,048VRAM: 41.4 GBHeadroom: 6.6 GBTTFT: noticeable
20
tok/s
E
Weights
32.00 GB
KV cache
4.00 GB
Activations
3.65 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~992 ms (noticeable)
Model details →Run-on benchmark page →
#12Phi-4 14B
14B
phi
Commercial OK
Quant: Q8_0Context: 8,192VRAM: 32.6 GBHeadroom: 15.4 GBTTFT: fast
ollama run phi4:14b
44
tok/s
E
Weights
14.88 GB
KV cache
7.00 GB
Activations
8.94 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~434 ms (fast)
Model details →Run-on benchmark page →

Runs with tradeoffs
22 models

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

DeepSeek R1 Distill Qwen 32B
32B
deepseek
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 46.3 GBHeadroom: 1.7 GBTTFT: noticeable
  • • Tight VRAM fit — only 1.7 GB headroom left for context growth
ollama run deepseek-r1:32b
34
tok/s
E
Weights
19.32 GB
KV cache
16.00 GB
Activations
9.16 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~992 ms (noticeable)
Model details →Run-on benchmark page →
QwQ 32B Preview
32B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 46.3 GBHeadroom: 1.7 GBTTFT: noticeable
  • • Tight VRAM fit — only 1.7 GB headroom left for context growth
ollama run qwq:32b
34
tok/s
E
Weights
19.32 GB
KV cache
16.00 GB
Activations
9.16 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~992 ms (noticeable)
Model details →Run-on benchmark page →
DeepSeek R1 Distill Llama 70B
70B
deepseek
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 57.0 GBHeadroom: 10.2 GBTTFT: slow
  • • Partial CPU offload: ~16% of layers run on CPU
ollama run deepseek-r1:70b
16
tok/s
E
Weights
42.26 GB
KV cache
8.75 GB
Activations
4.16 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~2169 ms (slow)
Model details →Run-on benchmark page →
Qwen 2.5 Math 72B
72B
qwen
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 58.5 GBHeadroom: 8.7 GBTTFT: slow
  • • Partial CPU offload: ~18% of layers run on CPU
15
tok/s
E
Weights
43.47 GB
KV cache
9.00 GB
Activations
4.22 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~2231 ms (slow)
Model details →Run-on benchmark page →
Nemotron 3 Nano (30B-A3B)
30B
other
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 44.0 GBHeadroom: 4.0 GBTTFT: noticeable
  • • Tight VRAM fit — only 4.0 GB headroom left for context growth
ollama run nemotron3:nano
36
tok/s
E
Weights
18.11 GB
KV cache
15.00 GB
Activations
9.10 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~930 ms (noticeable)
Model details →Run-on benchmark page →
Llama 3.1 Nemotron 70B Instruct
70B
llama
Commercial OK
Quant: Q4_K_MContext: 2,048VRAM: 57.0 GBHeadroom: 10.2 GBTTFT: slow
  • • Partial CPU offload: ~16% of layers run on CPU
ollama run nemotron:70b
16
tok/s
E
Weights
42.26 GB
KV cache
8.75 GB
Activations
4.16 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~2169 ms (slow)
Model details →Run-on benchmark page →
Nemotron 3 Super 49B
49B
other
Commercial OK
Quant: AWQ-INT4Context: 2,048VRAM: 61.4 GBHeadroom: 5.8 GBTTFT: noticeable
  • • Partial CPU offload: ~22% of layers run on CPU
13
tok/s
E
Weights
49.00 GB
KV cache
6.13 GB
Activations
4.50 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~1519 ms (noticeable)
Model details →Run-on benchmark page →
Qwen 3 30B-A3B
30B
qwen
Commercial OK
Quant: Q4_K_MContext: 8,192VRAM: 44.0 GBHeadroom: 4.0 GBTTFT: noticeable
  • • Tight VRAM fit — only 4.0 GB headroom left for context growth
ollama run qwen3:30b
36
tok/s
E
Weights
18.11 GB
KV cache
15.00 GB
Activations
9.10 GB
Runtime
1.80 GB
Time to first token (prefill, 512-token prompt): ~930 ms (noticeable)
Model details →Run-on benchmark page →

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: 36 new comfortable, 28 new tradeoff

  • • Gemma 3 1B
  • • Llama 3.2 1B Instruct
  • • Gemma 4 E2B (Effective 2B)
  • • Llama 3.2 3B Instruct
Shop this upgrade↗

Upgrade to NVIDIA H100 SXM

see current pricing

80 GB VRAM (vs your 48 GB) plus a bandwidth jump from ~1008 GB/s to ~3350 GB/s.

Unlocks: 59 new comfortable

  • • Gemma 3 1B
  • • Llama 3.2 1B Instruct
  • • Gemma 4 E2B (Effective 2B)
  • • Llama 3.2 3B Instruct
Shop this upgrade↗

Add a second NVIDIA RTX 4090 48GB (China-mod)

~$2400

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: 65 new comfortable

  • • Gemma 3 1B
  • • Llama 3.2 1B Instruct
  • • Gemma 4 E2B (Effective 2B)
  • • Llama 3.2 3B Instruct
Shop this upgrade↗

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 (48 GB) + 60% of system RAM (19 GB) combined.

—
Qwen 3.5 235B-A17B (MoE)
397B
qwen
Commercial OK

Even with CPU offload, needs more memory than your VRAM (48 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 (48 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 (48 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 (48 GB) + 60% of system RAM (19 GB) combined.

—

How to read these numbers

M
Measured — we ran this exact combo on owner hardware.

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

E
Estimated — pure formula based on VRAM bandwidth and model architecture.

Full methodology →

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