What GPU do I need to run LiquidAI/LFM2.5-350M-Base?
354M parameters, published in BF16. View on Hugging Face
Set up LFM2.5-350M-BaseLFM2.5-350M-Base is published by LiquidAI on Hugging Face, with 11,723 downloads and 32 likes to date. It's a Lfm2ForCausalLM model built for text-generation, published natively in BF16.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 0.7 GB | 0.8 GB | RTX 3060 | 1 | $0.110/hr |
| FP8 (quantized) | 0.3 GB | 0.4 GB | RTX 4000 SFF Ada | 1 | $0.198/hr |
| INT4 (quantized) | 0.2 GB | 0.2 GB | RTX 3060 | 1 | $0.110/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run LFM2.5-350M-Base at its published (BF16) precision: 1× RTX 3060, at $0.110/hr per GPU ($0.110/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
LFM2.5-350M-Base: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Hybrid: only the full-attention layers cache per token; the other layers keep a fixed-size state. Cached per token: 12,288 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.05 GB |
| 32K | 0.38 GB |
| 125K (model maximum) | 1.5 GB |
Computed from the layer, head and window counts in the model's own configuration (LiquidAI/LFM2.5-350M-Base), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
LFM2.5-350M-Base: common questions
How much VRAM does LFM2.5-350M-Base need?
0.8 GB at BF16, 0.4 GB at FP8 (quantized), 0.2 GB at INT4 (quantized). All of those sit under 6 GB, the smallest capacity this page reasons about, so VRAM is not what limits where this model runs. The 0.7 GB of weights plus inference overhead is the whole requirement.
How many copies of LFM2.5-350M-Base fit on one RTX 3060?
15, by VRAM alone. That card carries 12.0 GB and one copy needs 0.8 GB at BF16, on a live rate of $0.110/hr for the whole card. Throughput is not modelled here, so 15 copies is not 15 times the requests served.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More LFM2 models
- LFM2.5-350M (354M, BF16)
- LFM2.5-230M (230M, BF16)
- LFM2.5-230M-Base (230M, BF16)
- LFM2.5-1.2B-Instruct (1.2B, BF16)
- LFM2-1.2B (1.2B, BF16)
Lineage
Quantized and fine-tuned versions
Alternatives at this size
Other models for text-generation within about a third of LFM2.5-350M-Base's 354M parameters, from other model lines.
- Qwen2.5-0.5B-Instruct (494M, BF16)
- gemma-3-270m (268M, BF16)
- SmolLM2-360M (362M, BF16)
- gpt2-medium (380M, F32)
- pythia-410m (506M, F16)
More on LFM2.5-350M-Base
Fits on an 8 GB GPU at BF16: every model that fits in 8 GB.
Related reading: H100 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
