LFM2 2.6B

Liquid AI · liquid/lfm2-2-6b

ProviderLiquid AI
Familylfm2
Typellm-chat
Statusactive
Released2025-09-22
Updated2026-08-05
Parameters2.6B
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/sbf165.14 GB~2860732.2~11442928.7 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf165.14 GB~2179.6~8718.4 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf165.14 GB~1089.8~4359.2 q4
NVIDIA B200180.0 GB7700.0 GB/sbf165.14 GB~1048.9~4195.7 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf165.14 GB~1005.3~4021.4 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf165.14 GB~817.4~3269.4 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf165.14 GB~722.0~2888.0 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf165.14 GB~653.9~2615.5 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf165.14 GB~531.3~2125.1 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf165.14 GB~456.4~1825.4 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf165.14 GB~435.9~1743.7 q4
Google TPU v5p95.0 GB2765.0 GB/sbf165.14 GB~376.7~1506.7 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf165.14 GB~277.8~1111.1 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf165.14 GB~272.5~1089.8 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf165.14 GB~244.1~976.5 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf165.14 GB~223.1~892.5 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf165.14 GB~218.0~871.8 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf165.14 GB~211.8~847.3 q4
Google TPU v432.0 GB1200.0 GB/sbf165.14 GB~163.5~653.9 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf165.14 GB~137.3~549.3 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf165.14 GB~130.8~523.1 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sbf165.14 GB~130.8~523.1 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf165.14 GB~127.5~510.0 q4
NVIDIA L40S48.0 GB864.0 GB/sbf165.14 GB~117.7~470.8 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf165.14 GB~111.6~446.3 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf165.14 GB~109.0~435.9 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf165.14 GB~109.0~435.9 q4
Google TPU v5e16.0 GB800.0 GB/sbf165.14 GB~109.0~435.9 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sbf165.14 GB~100.3~401.0 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sbf165.14 GB~91.5~366.2 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sbf165.14 GB~87.2~348.7 q4
Apple M4 Max128.0 GB546.0 GB/sbf165.14 GB~74.4~297.5 q4
Apple M1 Max64.0 GB400.0 GB/sbf165.14 GB~54.5~218.0 q4
Apple M2 Max96.0 GB400.0 GB/sbf165.14 GB~54.5~218.0 q4
Apple M3 Max128.0 GB400.0 GB/sbf165.14 GB~54.5~218.0 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sbf165.14 GB~49.0~196.2 q4
NVIDIA L424.0 GB300.0 GB/sbf165.14 GB~40.9~163.5 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sbf165.14 GB~39.2~156.9 q4
Apple M4 Pro64.0 GB273.0 GB/sbf165.14 GB~37.2~148.8 q4
Apple M432.0 GB120.0 GB/sbf165.14 GB~16.3~65.4 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLiquidAI/LFM2-2.6B

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 2.6B0.0%2026-09-10 evaluatedindependent evaluatorsource
critptLFM2 2.6B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondLFM2 2.6B30.61%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