L
localai.computer
ModelsGPUsSystemsBuildsOpenClawMethodology

Resources

  • Methodology
  • Submit Benchmark
  • About

Browse

  • AI Models
  • GPUs
  • PC Builds
  • AI News

Guides

  • OpenClaw Guide
  • How-To Guides

Legal

  • Privacy
  • Terms
  • Contact

© 2026 localai.computer. Hardware recommendations for running AI models locally.

ℹ️We earn from qualifying purchases through affiliate links at no extra cost to you. This supports our free content and research.

  1. Home
  2. GPUs
  3. RTX 5090

RTX 5090

By NVIDIAReleased 2025-01Launch MSRP $1,999.00

This GPU offers reliable throughput for local AI workloads. Pair it with the right model quantization to hit your desired tokens/sec, and monitor prices below to catch the best deal.

Check Price on AmazonView Benchmarks
Specs snapshot
Key hardware metrics for AI workloads.
VRAM32GB
Cores21,760
TDP575W
ArchitectureBlackwell

Quick Answer: RTX 5090 has 32GB VRAM, enough for models up to roughly 80B parameters at 4-bit quantization. It draws 575W under load.

Key Takeaways
  • 32GB VRAM - runs models up to ~80B parameters
  • Flagship-class compute for maximum throughput
  • High power draw (575W) - requires robust PSU (850W+ recommended)
  • Strong price-to-VRAM value

What this means for you

With 32GB VRAM, RTX 5090 can run models up to approximately 80B parameters using 4-bit quantization. That covers most popular models, including 70B-class ones at 4-bit.

Who should buy

  • Running 70B parameter models at good speeds
  • Multiple model instances simultaneously
  • Production deployments on single GPU

Looking to upgrade?

Consider RTX 4090 or RTX 6000 Ada — 24GB Ada offers better efficiency than Ampere.

Where to Buy

Buy directly on Amazon with fast shipping and reliable customer service.

Amazon
See price on Amazon
Buy on Amazon

Prime shipping available • 30-day returns

Complete Your Build

Essential accessories to pair with RTX 5090

Corsair RM750x ATX 3.1 750W
Minimum 750W recommended for RTX 40 series
~$119
View on Amazon
Corsair Vengeance RGB 32GB DDR5-6000
32GB ideal for AI workloads
~$129
View on Amazon
Noctua NF-A12x25 PWM
Quiet and efficient cooling
~$35
View on Amazon
Thermal Grizzly Kryonaut
Premium thermal paste for optimal cooling
~$15
View on Amazon

Accessories Total

Typical prices — check Amazon for current pricing

~$298
See Complete BuildsMore GPUs

💡 Not ready to buy? Try cloud GPUs first

Test RTX 5090 performance in the cloud before investing in hardware. Pay by the hour with no commitment.

Vast.aifrom $0.20/hrRunPodfrom $0.30/hrLambda Labsenterprise-grade

AI benchmarks

Showing 12 of 80 rows. Speeds are calculated estimates, not measurements — search for your model to jump straight to it.

ModelSizeQuantizationTokens/secVRAM used
Deepseek AI Deepseek Coder 1.3B Instruct1.3BQ4
~450 tok/sEstimated
1GB
Deepseek AI Deepseek R1 Distill Qwen 1.5B1.5BQ4
~450 tok/sEstimated
1GB
Deepseek AI Deepseek Ocr 2UnknownQ4
~375 tok/sEstimated
2GB
Deepseek AI Deepseek OcrUnknownQ4
~375 tok/sEstimated
2GB
Lmstudio Community Deepseek R1 0528 Qwen3 8B Mlx 8bit8BQ4
~375 tok/sEstimated
4GB
Lmstudio Community Deepseek R1 0528 Qwen3 8B Mlx 4bit8BQ4
~375 tok/sEstimated
4GB
Deepseek AI Deepseek R1 Distill Qwen 7B7BQ4
~375 tok/sEstimated
4GB
Nineninesix Kani Tts 2 EnUnknownQ4
~360 tok/sEstimated
1GB
Qwen Qwen3 Tts 12hz 1 7B Customvoice7BQ4
~360 tok/sEstimated
1GB
Zai Org Glm OcrUnknownQ4
~360 tok/sEstimated
1GB
Qwen Qwen3 Asr 1 7B7BQ4
~360 tok/sEstimated
2GB
Nari Labs Dia2 2B2BQ4
~360 tok/sEstimated
1GB
Deepseek AI Deepseek Coder 1.3B Instruct
Q4 · 1.3B
1GB
~450 tok/sEstimated
Deepseek AI Deepseek R1 Distill Qwen 1.5B
Q4 · 1.5B
1GB
~450 tok/sEstimated
Deepseek AI Deepseek Ocr 2
Q4 · Unknown
2GB
~375 tok/sEstimated
Deepseek AI Deepseek Ocr
Q4 · Unknown
2GB
~375 tok/sEstimated
Lmstudio Community Deepseek R1 0528 Qwen3 8B Mlx 8bit
Q4 · 8B
4GB
~375 tok/sEstimated
Lmstudio Community Deepseek R1 0528 Qwen3 8B Mlx 4bit
Q4 · 8B
4GB
~375 tok/sEstimated
Deepseek AI Deepseek R1 Distill Qwen 7B
Q4 · 7B
4GB
~375 tok/sEstimated
Nineninesix Kani Tts 2 En
Q4 · Unknown
1GB
~360 tok/sEstimated
Qwen Qwen3 Tts 12hz 1 7B Customvoice
Q4 · 7B
1GB
~360 tok/sEstimated
Zai Org Glm Ocr
Q4 · Unknown
1GB
~360 tok/sEstimated
Qwen Qwen3 Asr 1 7B
Q4 · 7B
2GB
~360 tok/sEstimated
Nari Labs Dia2 2B
Q4 · 2B
1GB
~360 tok/sEstimated

Model compatibility

Showing 12 of 240 rows.

ModelSizeQuantizationVerdictEstimated speedVRAM needed
01 AI Yi 1 5 34B Chat34BQ4Fits comfortably
~105 tok/sEstimated
18GB (have 32GB)
01 AI Yi 1 5 34B Chat34BQ8Not supported
~73 tok/sEstimated
35GB (have 32GB)
01 AI Yi 1 5 34B Chat34BFP16Not supported
~40 tok/sEstimated
69GB (have 32GB)
AI Forever Rugpt 3.5 13B13BQ4Fits comfortably
~225 tok/sEstimated
7GB (have 32GB)
AI Forever Rugpt 3.5 13B13BQ8Fits comfortably
~155 tok/sEstimated
13GB (have 32GB)
AI Forever Rugpt 3.5 13B13BFP16Fits comfortably
~85 tok/sEstimated
26GB (have 32GB)
AI Mo Kimina Prover 72B72BQ4Not supported
~60 tok/sEstimated
37GB (have 32GB)
AI Mo Kimina Prover 72B72BQ8Not supported
~42 tok/sEstimated
73GB (have 32GB)
AI Mo Kimina Prover 72B72BFP16Not supported
~23 tok/sEstimated
146GB (have 32GB)
Alibaba Nlp Gte Qwen2 1.5B Instruct1.5BQ4Fits comfortably
~360 tok/sEstimated
1GB (have 32GB)
Alibaba Nlp Gte Qwen2 1.5B Instruct1.5BQ8Fits comfortably
~250 tok/sEstimated
2GB (have 32GB)
Alibaba Nlp Gte Qwen2 1.5B Instruct1.5BFP16Fits comfortably
~135 tok/sEstimated
4GB (have 32GB)
01 AI Yi 1 5 34B ChatQ4
Size: 34B
Fits comfortably18GB required · 32GB available
~105 tok/sEstimated
01 AI Yi 1 5 34B ChatQ8
Size: 34B
Not supported35GB required · 32GB available
~73 tok/sEstimated
01 AI Yi 1 5 34B ChatFP16
Size: 34B
Not supported69GB required · 32GB available
~40 tok/sEstimated
AI Forever Rugpt 3.5 13BQ4
Size: 13B
Fits comfortably7GB required · 32GB available
~225 tok/sEstimated
AI Forever Rugpt 3.5 13BQ8
Size: 13B
Fits comfortably13GB required · 32GB available
~155 tok/sEstimated
AI Forever Rugpt 3.5 13BFP16
Size: 13B
Fits comfortably26GB required · 32GB available
~85 tok/sEstimated
AI Mo Kimina Prover 72BQ4
Size: 72B
Not supported37GB required · 32GB available
~60 tok/sEstimated
AI Mo Kimina Prover 72BQ8
Size: 72B
Not supported73GB required · 32GB available
~42 tok/sEstimated
AI Mo Kimina Prover 72BFP16
Size: 72B
Not supported146GB required · 32GB available
~23 tok/sEstimated
Alibaba Nlp Gte Qwen2 1.5B InstructQ4
Size: 1.5B
Fits comfortably1GB required · 32GB available
~360 tok/sEstimated
Alibaba Nlp Gte Qwen2 1.5B InstructQ8
Size: 1.5B
Fits comfortably2GB required · 32GB available
~250 tok/sEstimated
Alibaba Nlp Gte Qwen2 1.5B InstructFP16
Size: 1.5B
Fits comfortably4GB required · 32GB available
~135 tok/sEstimated

Note: Performance estimates are calculated. Real results may vary. Methodology · Submit real data

Alternative GPUs

RTX 4090
24GB

Explore how RTX 4090 stacks up for local inference workloads.

RTX 4080
16GB

Explore how RTX 4080 stacks up for local inference workloads.

RTX 4070 Ti
12GB

Explore how RTX 4070 Ti stacks up for local inference workloads.

RTX 3090
24GB

Explore how RTX 3090 stacks up for local inference workloads.

RX 7900 XTX
24GB

Explore how RX 7900 XTX stacks up for local inference workloads.