Head to head · October 2026
RTX 4090 vs RTX 5090: which is better for local AI?
24GB against 32GB, 1,008 against 1,792 GB/s. What each runs, how fast, and which one to buy.
The verdict
The RTX 5090 has 8GB more memory (32GB vs 24GB), so it runs larger models. The RTX 5090 generates text about 60% faster thanks to 1,792 GB/s of memory bandwidth versus 1,008 GB/s. The RTX 4090 costs about $1,800 less ($2,500 used vs $4,300).
- Get the RTX 4090 if you want the most for your money or you care about power draw, heat, and noise.
- Get the RTX 5090 if you want to run bigger models or use longer context, speed matters most, or you want a new card with a warranty.
We may earn a commission when you buy through links on this page. It never changes what we recommend. How we make money
Performance
Model by model
Estimated single-user generation speed for popular models, at the best quantization each card can hold with 8K context.
| Model | RTX 4090 | RTX 5090 |
|---|---|---|
| Llama 3.1 8B Instruct | ~86 tok/sQ8_0 | ~145 tok/sQ8_0 |
| Qwen3 14B | ~49 tok/sQ8_0 | ~84 tok/sQ8_0 |
| gpt-oss-20b | ~135 tok/sMXFP4 | ~190 tok/sMXFP4 |
| Qwen3 30B-A3B (2507) | ~150 tok/sQ4_K_M | ~175 tok/sQ6_K |
| Gemma 3 27B | ~46 tok/sQ4_K_M | ~60 tok/sQ6_K |
| Qwen3 32B | ~48 tok/sQ3_K_M | ~58 tok/sQ5_K_M |
| Llama 3.3 70B Instruct | ~1.8 tok/sQ4_K_M Partial offload | ~2.7 tok/sQ4_K_M Partial offload |
| gpt-oss-120b | ~21 tok/sMXFP4 Partial offload | ~25 tok/sMXFP4 Partial offload |
Estimates from our bandwidth model; see methodology. Measured results, where we have them, are on each GPU's page.
Specifications
Spec by spec
| Spec | RTX 4090 | RTX 5090 |
|---|---|---|
| Memory | 24GB GDDR6X | 32GB GDDR7● |
| Bandwidth | 1,008 GB/s | 1,792 GB/s● |
| Power | 450W● | 575W |
| Price | ~$2,500 used● | ~$4,300 |
| Price per GB | $104● | $134 |
| Architecture | Ada Lovelace | Blackwell |
| Released | 2022 | 2025 |
| Software | CUDA, Vulkan | CUDA, Vulkan |
Our take
Bottom line on each
Still one of the fastest 24GB cards, though used prices now exceed its launch MSRP.
The fastest consumer card for local LLMs, but shortage pricing has more than doubled its cost.
Frequently asked questions
Is the RTX 4090 or RTX 5090 better for running LLMs?
The RTX 5090 has 8GB more memory (32GB vs 24GB), so it runs larger models. The RTX 5090 generates text about 60% faster thanks to 1,792 GB/s of memory bandwidth versus 1,008 GB/s. The RTX 4090 costs about $1,800 less ($2,500 used vs $4,300).
How much faster is the RTX 5090 than the RTX 4090 for AI?
About 60% faster at generating text in our estimates, because token generation is limited by memory bandwidth (1,792 GB/s vs 1,008 GB/s). Prompt processing depends more on compute and can differ by more.
Keep comparing