On an RTX 5070 Ti (16 GB), Gemma 2 2B in Q4_K_M needs about 3,1 Go of video memory: it should fit. Estimated generation speed: ≈ 344 tokens/s.
Not measured yet: the figures below are estimates, computed from the models already measured on this card. Measure my card
≈ 344 tok/s
Generation (estimated)
≥ 3,1 Go
Recommended VRAM (16 GB avail.)
≈ 103 W
Power (ref. model)
0,02 €/M tok
Electricity / M tokens · 0,25 €/kWh
Calculation: Qwen2.5 3B measured at 305 tok/s on this card, 1,8 Go read per token → × 1,8 Go ÷ 1,6 Go ≈ 344 tok/s (generation is memory-bound: speed follows the amount of weights read).
What about a longer context?
Context
KV cache
Estimated VRAM
Fits in 16 GB?
Estimated speed
4 096
0,2 Go
2,3 Go
yes
≈ 311 tok/s
8 192
0,4 Go
2,5 Go
yes
≈ 279 tok/s
Estimated VRAM = file + KV cache (full-attention layers, f16) + ~0,5 Go of buffers, excluding memory used by the system — hence the slightly higher “recommended VRAM”, which keeps that margin. Speed ≈ estimated speed × weights read ÷ (weights read + added KV cache).
Estimates based on real measurements of other models on this card and on the model's public data (Google Gemma). As soon as a measurement exists, this page shows the real figures.