GeForce RTX 4080
A 16 GB desktop card: how much local model fits and how fast it runs.
03 · AI Reviews
Local LLM inference on real hardware. Every figure is published with the GPU driver and runtime it was measured on, so you can reproduce it.
A 16 GB desktop card: how much local model fits and how fast it runs.
The top laptop GPU for local AI, where power limits and cooling matter as much as the silicon.
A 24 GB Blackwell workstation card at the entry of the PRO line.
The middle Blackwell workstation card, with 32 GB for larger models.
A 48 GB Blackwell workstation card aimed at the biggest local models.
No benchmarks are published yet. Results appear here, side by side, as each GPU is tested.
Each review states whether the hardware was bought, loaned or sponsored. Results from different drivers or runtimes are not directly comparable, which is why both are listed.