LFM2 1.2B

Liquid AI · liquid/lfm2-1-2b

ProviderLiquid AI
Familylfm2
Typellm-chat
Statusactive
Released2025-07-10
Updated2026-08-05
Parameters1.2B
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.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
Cerebras WSE-344.0 GB21000000.0 GB/sbf162.34 GB~6280223.0~25120892.0 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf162.34 GB~4784.9~19139.7 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf162.34 GB~2392.5~9569.9 q4
NVIDIA B200180.0 GB7700.0 GB/sbf162.34 GB~2302.7~9211.0 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf162.34 GB~2207.0~8828.2 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf162.34 GB~1794.3~7177.4 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf162.34 GB~1585.0~6340.0 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf162.34 GB~1435.5~5741.9 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf162.34 GB~1166.3~4665.3 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf162.34 GB~1001.8~4007.4 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf162.34 GB~957.0~3827.9 q4
Google TPU v5p95.0 GB2765.0 GB/sbf162.34 GB~826.9~3307.6 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf162.34 GB~609.8~2439.1 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf162.34 GB~598.1~2392.5 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf162.34 GB~535.9~2143.6 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf162.34 GB~489.9~1959.4 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf162.34 GB~478.5~1914.0 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf162.34 GB~465.0~1860.1 q4
Google TPU v432.0 GB1200.0 GB/sbf162.34 GB~358.9~1435.5 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf162.34 GB~301.5~1205.8 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf162.34 GB~287.1~1148.4 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sbf162.34 GB~287.1~1148.4 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf162.34 GB~279.9~1119.7 q4
NVIDIA L40S48.0 GB864.0 GB/sbf162.34 GB~258.4~1033.5 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf162.34 GB~244.9~979.7 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf162.34 GB~239.2~957.0 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf162.34 GB~239.2~957.0 q4
Google TPU v5e16.0 GB800.0 GB/sbf162.34 GB~239.2~957.0 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sbf162.34 GB~220.1~880.4 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sbf162.34 GB~201.0~803.9 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sbf162.34 GB~191.4~765.6 q4
Apple M4 Max128.0 GB546.0 GB/sbf162.34 GB~163.3~653.1 q4
Apple M1 Max64.0 GB400.0 GB/sbf162.34 GB~119.6~478.5 q4
Apple M2 Max96.0 GB400.0 GB/sbf162.34 GB~119.6~478.5 q4
Apple M3 Max128.0 GB400.0 GB/sbf162.34 GB~119.6~478.5 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sbf162.34 GB~107.7~430.6 q4
NVIDIA L424.0 GB300.0 GB/sbf162.34 GB~89.7~358.9 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sbf162.34 GB~86.1~344.5 q4
Apple M4 Pro64.0 GB273.0 GB/sbf162.34 GB~81.6~326.6 q4
Apple M432.0 GB120.0 GB/sbf162.34 GB~35.9~143.5 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLiquidAI/LFM2-1.2B

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 1.2B0.0%2026-09-10 evaluatedindependent evaluatorsource
critptLFM2 1.2B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondLFM2 1.2B22.83%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