What GPU do you need to run LFM3-1.7B?

1.73B parameters (Hugging Face, accessed 2026-10-27). Q4_K_M GGUF: 1.1 GB (source). Architecture config (Hugging Face config.json; layer_types: 6 full_attention + 5 sliding_attention (256) + 13 conv layers. The conv layers keep a small fixed-size state and no KV (arch hybrid, so the speed estimate adds the recurrent-state cost); llama.cpp lfm3 puts the attention layers in its iSWA cache (src/models/lfm3.cpp, llama-memory-hybrid-iswa)): 6 attention layers (the rest are recurrent) × 4 KV heads × 128 head dim, plus 5 sliding-window layers (256-token window), so about 12 KB of FP16 KV cache per token of context (how we estimate context).

Liquid AI text model on the LFM3 backbone: 24 layers, 13 gated short convolutions (a small fixed-size state, no growing KV cache), 5 sliding-window attention layers (256-token window) and 6 global attention layers. 1.73B is Hugging Face's safetensors total of 1,728,801,536 (BF16). Card: 256K context; reasoning is on by default. Q4_K_M is Liquid AI's official GGUF (LFM3-1.7B-Q4_K_M.gguf, 1,099,627,808 bytes); the repo also has a quantization-aware-distilled QAD-Q4_0 build. Every size is in the table below. llama.cpp support (lfm3 architecture) was merged on Oct 27, 2026 (ggml-org/llama.cpp PR #31252) and is in tagged releases v0.9.0 (Nov 4, 2026) and v1.0.0 (Nov 13, 2026). Liquid's MLX builds wait on an mlx-lm release with LFM3 support (in progress, per the card); Transformers needs trust_remote_code. Liquid says the model can power agentic applications that used to need 9B parameters and that a 128K-token context needs 1.6 GB of KV cache; those are Liquid's claims, not checked here. LFM Open License v1.0: commercial use is licensed only while you (with your affiliates) stay under $10M annual revenue; above that, commercial use is not licensed under it. The cap does not apply to qualifying non-profits doing non-commercial or research work. Release date is Liquid's announcement.