Measured on our own hardware

The 16 GB Local AI Playbook

Every number here is measured on one real card — an NVIDIA GeForce RTX 5060 Ti with 16 GB of memory — not estimated from a spec sheet. We own exactly one consumer GPU, so instead of guessing at hardware we don't have, we go deep on the tier most people actually run: what fits in 16 GB, at what speed, at the quant a buyer would use.

Last measured August 4, 2026 · driver 580.173.02 · Ollama 0.23.2

What fits your card?

Speed is measured on the 16 GB card above. Whether a modelfits is mostly its memory footprint, which travels between cards far better than its speed does — so filter by the VRAM you have and see which of these measured models fit inside it.

Filter by VRAM

The measured ladder

ModelParamsQuantVRAM usedOn GPUGenerationPromptCold load
Gemma 3 · 1Bollama pull gemma3:1b999.89MQ4_K_M1.3 GB100% GPU265 tok/s7996 tok/s1.73s
Qwen2.5 · 3B Instructollama pull qwen2.5:3b-instruct3.1BQ4_K_M3.1 GB100% GPU176 tok/s9761 tok/s1.59s
Gemma 3 · 4Bollama pull gemma3:4b4.3BQ4_K_M4.6 GB100% GPU120 tok/s3886 tok/s3.04s
Llama 3.2 · 3Bollama pull llama3.2:3b3.2BQ4_K_M4.8 GB100% GPU175 tok/s9020 tok/s1.78s
Qwen2.5 · 7B Instructollama pull qwen2.5:7b-instruct7.6BQ4_K_M6.3 GB100% GPU88 tok/s4956 tok/s2.14s
Qwen2.5 · 14Bollama pull qwen2.5:14b14.8BQ4_K_M13 GB100% GPU45 tok/s2549 tok/s3.69s

How these were measured

  • One NVIDIA GeForce RTX 5060 Ti (16 GB). Generation rate is eval tokens ÷ eval time from the Ollama API, taken as the median of 3 timed runs of 200 tokens after a warm-up.
  • VRAM used and “on GPU” are what ollama psreports for the resident model — the model's real footprint, and whether all of it sits on the GPU or spills to the CPU (which is the moment speed falls off a cliff).
  • Params, quantisation and architecture are read fromollama show — the model file's own metadata, not typed by us. Everything shown is Q4_K_M-class weights, the quant most people actually run.
  • Speed is specific to this card; your tokens/sec will differ on other hardware. Footprint and fit travel much better — that's why the filter above keys off memory, not speed.