RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run LFM2.5 8B A1B locally?

LFM2.5 8B A1B by LiquidAI has 8.5B parameters (1.0B active per token, 32 experts, 4 used per token). At 4-bit with an 8K context it needs about 5.5 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.

5.5 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
125Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from LiquidAI/LFM2.5-8B-A1B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx
2-bit (Q2_K)rough3.6 GB3.7 GB4.0 GB
3-bit (Q3_K_M)usable4.5 GB4.6 GB4.9 GB
4-bit (Q4_K_M) •good, most popular5.5 GB5.5 GB5.8 GB
5-bit (Q5_K_M) •very good6.3 GB6.4 GB6.6 GB
6-bit (Q6_K) •near lossless7.2 GB7.2 GB7.5 GB
8-bit (Q8_0) •effectively lossless9.1 GB9.1 GB9.4 GB
16-bit (F16/BF16) •original weights32.3 GB32.3 GB32.6 GB

Which GPUs and Macs run LFM2.5 8B A1B?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsQ8_09.1 GB338details
RTX 4060 Ti 16GB16 GBFitsQ8_09.1 GB271details
RTX 309024 GBFitsQ8_09.1 GB880details
RTX 409024 GBFitsQ8_09.1 GB947details
RTX 509032 GBFitsQ8_09.1 GB1684details
RTX 5070 Ti16 GBFitsQ8_09.1 GB842details
RX 7900 XTX24 GBFitsQ8_09.1 GB902details
Mac M4 Pro 48GB48 GBFitsQ8_09.1 GB257details
Mac M4 Max 128GB128 GBFitsF1632.3 GB136details
Mac M3 Ultra 512GB512 GBFitsF1632.3 GB205details
DGX Spark128 GBFitsF1632.3 GB68details
RTX 407012 GBFitsQ8_09.1 GB474details
RTX 5060 Ti 16GB16 GBFitsQ8_09.1 GB421details
RTX 40608 GBFitsQ5_K_M6.4 GB382details
Arc B58012 GBFitsQ8_09.1 GB429details
RTX 3060 Ti8 GBFitsQ5_K_M6.4 GB629details
RTX 30708 GBFitsQ5_K_M6.4 GB629details
RTX 3080 10GB10 GBFitsQ6_K7.2 GB925details
RTX 3080 12GB12 GBFitsQ8_09.1 GB857details
RTX 3090 Ti24 GBFitsQ8_09.1 GB947details
RTX 4060 Ti 8GB8 GBFitsQ5_K_M6.4 GB404details
RTX 4070 Super12 GBFitsQ8_09.1 GB474details
RTX 4070 Ti12 GBFitsQ8_09.1 GB474details
RTX 4070 Ti Super16 GBFitsQ8_09.1 GB632details
RTX 408016 GBFitsQ8_09.1 GB674details
RTX 4080 Super16 GBFitsQ8_09.1 GB692details
RTX 50608 GBFitsQ5_K_M6.4 GB629details
RTX 507012 GBFitsQ8_09.1 GB632details
RTX 508016 GBFitsQ8_09.1 GB902details
2x RTX 30902×24 GBFitsF1632.3 GB234details
2x RTX 40902×24 GBFitsF1632.3 GB252details
2x RTX 50902×32 GBFitsF1632.3 GB448details
RTX A600048 GBFitsF1632.3 GB192details
RTX 6000 Ada48 GBFitsF1632.3 GB240details
RTX PRO 6000 Blackwell96 GBFitsF1632.3 GB448details
L40S48 GBFitsF1632.3 GB216details
A100 40GB40 GBFitsF1632.3 GB388details
A100 80GB80 GBFitsF1632.3 GB509details
H100 PCIe80 GBFitsF1632.3 GB500details
H100 SXM80 GBFitsF1632.3 GB837details
H200141 GBFitsF1632.3 GB1199details
RX 76008 GBFitsQ5_K_M6.4 GB404details
RX 7700 XT12 GBFitsQ8_09.1 GB406details
RX 7800 XT16 GBFitsQ8_09.1 GB586details
RX 7900 GRE16 GBFitsQ8_09.1 GB541details
RX 7900 XT20 GBFitsQ8_09.1 GB752details
RX 907016 GBFitsQ8_09.1 GB601details
RX 9070 XT16 GBFitsQ8_09.1 GB601details
Ryzen AI Max+ 395 128GB128 GBFitsF1632.3 GB64details
Arc A770 16GB16 GBFitsQ8_09.1 GB526details
Mac M1 16GB16 GBFitsQ8_09.1 GB64details
Mac M2 24GB24 GBFitsQ8_09.1 GB94details
Mac M3 24GB24 GBFitsQ8_09.1 GB94details
Mac M4 16GB16 GBFitsQ8_09.1 GB113details
Mac M4 32GB32 GBFitsQ8_09.1 GB113details
Mac M1 Pro 32GB32 GBFitsQ8_09.1 GB188details
Mac M3 Pro 36GB36 GBFitsQ8_09.1 GB141details
Mac M4 Pro 24GB24 GBFitsQ8_09.1 GB257details
Mac M4 Pro 64GB64 GBFitsF1632.3 GB68details
Mac M1 Max 64GB64 GBFitsF1632.3 GB100details
Mac M2 Max 96GB96 GBFitsF1632.3 GB100details
Mac M3 Max 128GB128 GBFitsF1632.3 GB100details
Mac M4 Max 64GB64 GBFitsF1632.3 GB136details
Mac M2 Ultra 192GB192 GBFitsF1632.3 GB200details
Mac M3 Ultra 256GB256 GBFitsF1632.3 GB205details

Questions

How much VRAM does LFM2.5 8B A1B need?

About 5.5 GB at 4-bit (Q4_K_M) with an 8K context: 4.8 GB of weights (real GGUF file size), 0.1 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 9.1 GB.

What is the smallest GPU that runs LFM2.5 8B A1B?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).

Can I run LFM2.5 8B A1B on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB or larger, using the default macOS GPU memory limit.

Similar models

Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.

d1 3B

3.1B · 2.3 GB at 4-bit

LFM2.5 VL 3B

3.1B · 2.3 GB at 4-bit

LFM2.5 2.6B

2.7B · 2.3 GB at 4-bit

Qwen2.5 VL 7B Instruct

8.3B · 5.5 GB at 4-bit

UI TARS 1.5 7B

8.3B · 5.5 GB at 4-bit

olmOCR 2 7B 1025

8.3B · 5.5 GB at 4-bit

Model facts

Hugging Face repoLiquidAI/LFM2.5-8B-A1B
Parameters8,467,856,832 (8.5B) · ~1.0B active (from the model name)
Layers / hidden size24 / 2048
AttentionGQA (hybrid: 18 linear/state layers); 32 heads, 8 KV heads, head dim 64
KV cache per token (FP16)12.0 KiB
Max context128,000 tokens
Licenseother
Released on HFMay 28, 2026
Downloads (30d) / likes16,577 / 808

Sources: Hugging Face model API · config.json (LiquidAI/LFM2.5-8B-A1B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.