Inference · compare

Gemma 4 E2B vs Qwen 3 4B.

Side by side only on machines of the same kind, so the difference is the model, not the computer. Each figure is the fastest build of that model on that machine.

4machine kinds both ran on
MachineGemma 4 E2B tok/sQwen 3 4B tok/sContext A / B
Linux, 4 CPUs · 31 GB84Q4_0 (4-bit) by unsloth61IQ1_S (1-bit) by unsloth128K / —
Windows, 8 CPUs · 16 GB13Q3_K_S (3-bit) by unsloth4.6IQ3_XXS (3-bit) by unsloth4K / 32K
Mac, 10 CPUs · 16 GB58Q4_0 (4-bit) by unsloth23Q5_0 (5-bit) by Qwen4K / 4K
Windows, 16 CPUs · 16 GB170Q4_K_M (4-bit) by google90IQ1_S (1-bit) by unsloth128K / —