RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run LFM2.5 24B A2B locally?

LFM2.5 24B A2B by LiquidAI has 24B parameters (2.0B active per token, 64 experts, 4 used per token). At 4-bit with an 8K context it needs about 14.3 GB of GPU memory. It fits on 59 of 65 GPUs and Macs we track. See all models that fit in 16 GB of VRAM.

14.3 GBQ4_K_M · 8K context
RX 7900 XTsmallest consumer GPU that fits
M2 24GBsmallest Mac that fits
128Kmax 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/LFM2.5-24B-A2B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K)rough9.1 GB9.2 GB9.6 GB11.5 GB
3-bit (Q3_K_M)usable11.6 GB11.7 GB12.2 GB14.1 GB
4-bit (Q4_K_M) •good, most popular14.3 GB14.3 GB14.8 GB16.7 GB
5-bit (Q5_K_M) •very good16.6 GB16.7 GB17.1 GB19.0 GB
6-bit (Q6_K) •near lossless19.1 GB19.1 GB19.6 GB21.5 GB
8-bit (Q8_0) •effectively lossless24.5 GB24.5 GB25.0 GB26.9 GB
16-bit (F16/BF16) •original weights90.1 GB90.1 GB90.6 GB92.5 GB

Which GPUs and Macs run LFM2.5 24B A2B?

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 GBFitsQ2_K9.2 GB480details
RTX 4060 Ti 16GB16 GBFitsQ3_K_M11.7 GB295details
RTX 309024 GBFitsQ6_K19.1 GB570details
RTX 409024 GBFitsQ6_K19.1 GB614details
RTX 509032 GBFitsQ8_024.5 GB843details
RTX 5070 Ti16 GBFitsQ3_K_M11.7 GB917details
RX 7900 XTX24 GBFitsQ6_K19.1 GB584details
Mac M4 Pro 48GB48 GBFitsQ8_024.5 GB128details
Mac M4 Max 128GB128 GBFitsQ8_024.5 GB257details
Mac M3 Ultra 512GB512 GBFitsF1690.1 GB102details
DGX Spark128 GBFitsF1690.1 GB34details
RTX 407012 GBFitsQ2_K9.2 GB672details
RTX 5060 Ti 16GB16 GBFitsQ3_K_M11.7 GB458details
RTX 40608 GBNo———details
Arc B58012 GBFitsQ2_K9.2 GB608details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBTightQ2_K9.2 GB1013details
RTX 3080 12GB12 GBFitsQ2_K9.2 GB1216details
RTX 3090 Ti24 GBFitsQ6_K19.1 GB614details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBFitsQ2_K9.2 GB672details
RTX 4070 Ti12 GBFitsQ2_K9.2 GB672details
RTX 4070 Ti Super16 GBFitsQ3_K_M11.7 GB687details
RTX 408016 GBFitsQ3_K_M11.7 GB734details
RTX 4080 Super16 GBFitsQ3_K_M11.7 GB753details
RTX 50608 GBNo———details
RTX 507012 GBFitsQ2_K9.2 GB896details
RTX 508016 GBFitsQ3_K_M11.7 GB982details
2x RTX 30902×24 GBFitsQ8_024.5 GB440details
2x RTX 40902×24 GBFitsQ8_024.5 GB474details
2x RTX 50902×32 GBFitsQ8_024.5 GB843details
RTX A600048 GBFitsQ8_024.5 GB361details
RTX 6000 Ada48 GBFitsQ8_024.5 GB451details
RTX PRO 6000 Blackwell96 GBFitsQ8_024.5 GB843details
L40S48 GBFitsQ8_024.5 GB406details
A100 40GB40 GBFitsQ8_024.5 GB731details
A100 80GB80 GBFitsQ8_024.5 GB959details
H100 PCIe80 GBFitsQ8_024.5 GB940details
H100 SXM80 GBFitsQ8_024.5 GB1575details
H200141 GBFitsF1690.1 GB600details
RX 76008 GBNo———details
RX 7700 XT12 GBFitsQ2_K9.2 GB576details
RX 7800 XT16 GBFitsQ3_K_M11.7 GB638details
RX 7900 GRE16 GBFitsQ3_K_M11.7 GB589details
RX 7900 XT20 GBFitsQ5_K_M16.7 GB563details
RX 907016 GBFitsQ3_K_M11.7 GB655details
RX 9070 XT16 GBFitsQ3_K_M11.7 GB655details
Ryzen AI Max+ 395 128GB128 GBFitsF1690.1 GB32details
Arc A770 16GB16 GBFitsQ3_K_M11.7 GB573details
Mac M1 16GB16 GBFitsQ2_K9.2 GB91details
Mac M2 24GB24 GBFitsQ4_K_M14.3 GB83details
Mac M3 24GB24 GBFitsQ4_K_M14.3 GB83details
Mac M4 16GB16 GBFitsQ2_K9.2 GB160details
Mac M4 32GB32 GBFitsQ6_K19.1 GB73details
Mac M1 Pro 32GB32 GBFitsQ6_K19.1 GB122details
Mac M3 Pro 36GB36 GBFitsQ6_K19.1 GB91details
Mac M4 Pro 24GB24 GBFitsQ4_K_M14.3 GB225details
Mac M4 Pro 64GB64 GBFitsQ8_024.5 GB128details
Mac M1 Max 64GB64 GBFitsQ8_024.5 GB188details
Mac M2 Max 96GB96 GBFitsQ8_024.5 GB188details
Mac M3 Max 128GB128 GBFitsQ8_024.5 GB188details
Mac M4 Max 64GB64 GBFitsQ8_024.5 GB257details
Mac M2 Ultra 192GB192 GBFitsF1690.1 GB100details
Mac M3 Ultra 256GB256 GBFitsF1690.1 GB102details

Prefer an API?

LiquidAI: LFM2.5-24B-A2B is on OpenRouter at $0.05 per 1M input tokens and $0.70 per 1M output tokens. Price details and history →

Questions

How much VRAM does LFM2.5 24B A2B need?

About 14.3 GB at 4-bit (Q4_K_M) with an 8K context: 13.5 GB of weights (real GGUF file size), 0.2 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 24.5 GB.

What is the smallest GPU that runs LFM2.5 24B A2B?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RX 7900 XT (20 GB).

Can I run LFM2.5 24B A2B on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M2 24GB 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.

LFM2.5 8B A1B

8.5B · 5.5 GB at 4-bit

d1 3B

3.1B · 2.3 GB at 4-bit

LFM2.5 VL 3B

3.1B · 2.3 GB at 4-bit

gemma 4 26B A4B it

26B · 17.2 GB at 4-bit

diffusiongemma 26B A4B it

26B · 17.1 GB at 4-bit

ERNIE 4.5 21B A3B PT

22B · 13.5 GB at 4-bit

Model facts

Hugging Face repoLiquidAI/LFM2.5-24B-A2B
Parameters23,971,585,280 (24B) · ~2.0B active (from the model name)
Layers / hidden size40 / 2048
AttentionGQA (hybrid: 30 linear/state layers); 32 heads, 8 KV heads, head dim 64
KV cache per token (FP16)20.0 KiB
Max context131,072 tokens
Licenseother
Released on HFOct 12, 2026
Downloads (30d) / likes31,046 / 618

Sources: Hugging Face model API · config.json (LiquidAI/LFM2.5-24B-A2B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.