Card × model estimate

LFM2.5 2.6B on RTX 4070 Ti Q4_K_M

should fit

On an RTX 4070 Ti (12 GB), LFM2.5 2.6B in Q4_K_M needs about 3,1 Go of video memory: it should fit. Estimated generation speed: ≈ 205 tokens/s.

Not measured yet: the figures below are estimates, computed from the models already measured on this card. Measure my card

≈ 205 tok/s

Generation (estimated)

≥ 3,1 Go

Recommended VRAM (12 GB avail.)

≈ 104 W

Power (ref. model)

0,04 €/M tok

Electricity / M tokens · 0,25 €/kWh

Calculation: Gemma 2 2B measured at 202 tok/s on this card, 1,6 Go read per token → × 1,6 Go ÷ 1,6 Go ≈ 205 tok/s (generation is memory-bound: speed follows the amount of weights read).

What about a longer context?

ContextKV cacheEstimated VRAMFits in 12 GB?Estimated speed
4 0960,1 Go2,1 Goyes≈ 199 tok/s
8 1920,1 Go2,2 Goyes≈ 191 tok/s
16 3840,3 Go2,3 Goyes≈ 178 tok/s
32 7680,5 Go2,6 Goyes≈ 156 tok/s
65 5361,0 Go3,1 Goyes≈ 126 tok/s
131 0722,0 Go4,1 Goyes≈ 90,5 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).

LFM2.5 2.6B on other cards

should fit
LFM2.5 2.6B pageRTX 4070 Ti pageFeel the speed Download the GGUF (1,6 Go)

Estimates based on real measurements of other models on this card and on the model's public data (Petits modèles). As soon as a measurement exists, this page shows the real figures.