RTX 4090 vs RTX 5090 for local LLMs
| RTX 4090 | RTX 5090 | |
|---|---|---|
| VRAM | 24 GB | 32 GB |
| Memory bandwidth | 1008 GB/s | 1792 GB/s |
| Lowest new price DE | ≈€3,955 | ≈€6,199 |
Measured decode speed, 2026 models (tok/s)
| Model | RTX 4090 | RTX 5090 | Difference |
|---|---|---|---|
| Gemma 4 12B | 98.1 | 142 | +45% |
| Gemma 4 26B-A4B | 164 | 235 | +43% |
| Gemma 4 31B | 43.8 | 70.0 | +60% |
| LFM2.5 24B-A2B | 261-339 | – | |
| Nemotron 3.5 Bolt 74B-A9B | 108 | – | |
| Qwen3.5 9B | 137 | 208 | +52% |
| Qwen3.6 27B | 47.0 | 75.7 | +61% |
| Qwen3.6 35B-A3B | – | 258 | |
| gpt-oss 20B | 203 | 313 | +54% |
Sources: RTX 4090 results, RTX 5090 results (community llama-bench via llamabench.ai; quants per result may differ, so treat differences as indicative). Specs per llamabench.ai; prices billiger.de 2026-10-16.
Not sure which you need? Check fit in the VRAM calculator or read best GPU for local LLMs in 2026.
Not enough VRAM? Rent before you buy. Run the same model on a RTX 5090 on a rented 24–96 GB cloud GPU by the hour and see the real speed before spending €1,000+. Try it on RunPod
Referral link: RigBase may earn RunPod credit if you sign up. It never affects our data.
Referral link: RigBase may earn RunPod credit if you sign up. It never affects our data.
SPONSOR
Your brand here. One sponsor slot on this page, shown to people choosing AI hardware. Sponsor RigBase →
Your brand here. One sponsor slot on this page, shown to people choosing AI hardware. Sponsor RigBase →