LFM2 24B A2B

Liquid AI · liquid/lfm2-24b-a2b

ProviderLiquid AI
Familylfm2
Typellm-chat
Statusactive
Released2026-02-24
Updated2026-08-05
Parameters23.8B
Open weightsyes

What it fits on

Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.

Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices. This is a mixture-of-experts model and no card carries active-parameter counts, so these predictions use total parameters and understate the real speed.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
Cerebras WSE-344.0 GB21000000.0 GB/sfp823.84 GB~616516.1~1233032.1 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1647.69 GB~234.9~939.5 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1647.69 GB~117.4~469.7 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1647.69 GB~113.0~452.1 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1647.69 GB~108.3~433.3 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1647.69 GB~88.1~352.3 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1647.69 GB~77.8~311.2 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1647.69 GB~70.5~281.8 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1647.69 GB~57.2~229.0 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf1647.69 GB~49.2~196.7 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1647.69 GB~47.0~187.9 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1647.69 GB~40.6~162.3 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf1647.69 GB~29.9~119.7 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf1647.69 GB~29.4~117.4 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sfp823.84 GB~52.6~105.2 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sfp823.84 GB~48.1~96.2 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf1647.69 GB~23.5~93.9 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sfp823.84 GB~45.7~91.3 q4
Google TPU v432.0 GB1200.0 GB/sfp823.84 GB~35.2~70.5 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sq617.88 GB~39.5~59.2 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sq617.88 GB~37.6~56.4 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sq411.92 GB~56.4~56.4 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sq617.88 GB~36.6~55.0 q4
NVIDIA L40S48.0 GB864.0 GB/sfp823.84 GB~25.4~50.7 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1647.69 GB~12.0~48.1 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sq514.9 GB~37.6~47.0 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1647.69 GB~11.7~47.0 q4
Google TPU v5e16.0 GB800.0 GB/sq411.92 GB~47.0~47.0 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sq411.92 GB~43.2~43.2 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sq411.92 GB~39.5~39.5 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sq411.92 GB~37.6~37.6 q4
Apple M4 Max128.0 GB546.0 GB/sbf1647.69 GB~8.0~32.1 q4
Apple M1 Max64.0 GB400.0 GB/sbf1647.69 GB~5.9~23.5 q4
Apple M2 Max96.0 GB400.0 GB/sbf1647.69 GB~5.9~23.5 q4
Apple M3 Max128.0 GB400.0 GB/sbf1647.69 GB~5.9~23.5 q4
NVIDIA L424.0 GB300.0 GB/sq617.88 GB~11.7~17.6 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sq411.92 GB~16.9~16.9 q4
Apple M4 Pro64.0 GB273.0 GB/sbf1647.69 GB~4.0~16.0 q4
Apple M432.0 GB120.0 GB/sfp823.84 GB~3.5~7.0 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLiquidAI/LFM2-24B-A2B

Verified benchmark evidence

Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.
BenchmarkEvaluated asScoreEvidence dateSource kind
aa_lcrLFM2 24B A2B0.0%2026-09-10 evaluatedindependent evaluatorsource
critptLFM2 24B A2B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondLFM2 24B A2B47.37%2026-09-10 evaluatedindependent evaluatorsource

Reported benchmark scores

This card reports no benchmark scores yet.

BenchmarkCatalogue standingScore

Data

This card as JSON · See it in the graph · Edit on GitHub