LFM2.5-2.6B now runs locally on your Mac ๐
Great release by
@LiquidAI โ and Nativ supports it Day 0.
Full BF16, no quantization:
โก๏ธ 11,000+ tok/s prefill
โก๏ธ 82 tok/s decode
๐ง Under 8.5GB peak memory โ even at full 128K context
๐ Scales to 476 tok/s aggregate across 16 concurrent requests
Try it on Nativ ๐
Github repo ๐ท