LFM2.5-2.6B

Liquid AI · liquid/lfm2-5-2-6b

ProviderLiquid AI
Familylfm2
Typellm-chat
Statusactive
Released2026-07-28
Updated2026-08-24
Parameters2.7B
Open weightsyes

Lineage

RelationshipModel
Is a finetune ofLiquidAI/LFM2.5-2.6B-Base

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/sbf165.39 GB~2725049.6~10900198.5 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf165.39 GB~2076.2~8304.9 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf165.39 GB~1038.1~4152.5 q4
NVIDIA B200180.0 GB7700.0 GB/sbf165.39 GB~999.2~3996.7 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf165.39 GB~957.7~3830.6 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf165.39 GB~778.6~3114.3 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf165.39 GB~687.8~2751.0 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf165.39 GB~622.9~2491.5 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf165.39 GB~506.1~2024.3 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf165.39 GB~434.7~1738.8 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf165.39 GB~415.2~1661.0 q4
Google TPU v5p95.0 GB2765.0 GB/sbf165.39 GB~358.8~1435.2 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf165.39 GB~264.6~1058.4 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf165.39 GB~259.5~1038.1 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf165.39 GB~232.5~930.2 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf165.39 GB~212.6~850.2 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf165.39 GB~207.6~830.5 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf165.39 GB~201.8~807.1 q4
Google TPU v432.0 GB1200.0 GB/sbf165.39 GB~155.7~622.9 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf165.39 GB~130.8~523.2 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf165.39 GB~124.6~498.3 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sbf165.39 GB~124.6~498.3 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf165.39 GB~121.5~485.8 q4
NVIDIA L40S48.0 GB864.0 GB/sbf165.39 GB~112.1~448.5 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf165.39 GB~106.3~425.1 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf165.39 GB~103.8~415.2 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf165.39 GB~103.8~415.2 q4
Google TPU v5e16.0 GB800.0 GB/sbf165.39 GB~103.8~415.2 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sbf165.39 GB~95.5~382.0 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sbf165.39 GB~87.2~348.8 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sbf165.39 GB~83.0~332.2 q4
Apple M4 Max128.0 GB546.0 GB/sbf165.39 GB~70.9~283.4 q4
Apple M1 Max64.0 GB400.0 GB/sbf165.39 GB~51.9~207.6 q4
Apple M2 Max96.0 GB400.0 GB/sbf165.39 GB~51.9~207.6 q4
Apple M3 Max128.0 GB400.0 GB/sbf165.39 GB~51.9~207.6 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sbf165.39 GB~46.7~186.9 q4
NVIDIA L424.0 GB300.0 GB/sbf165.39 GB~38.9~155.7 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sbf165.39 GB~37.4~149.5 q4
Apple M4 Pro64.0 GB273.0 GB/sbf165.39 GB~35.4~141.7 q4
Apple M432.0 GB120.0 GB/sbf165.39 GB~15.6~62.3 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLiquidAI/LFM2.5-2.6B

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_lcrLFM2.5-2.6B5.67%2026-09-10 evaluatedindependent evaluatorsource
critptLFM2.5-2.6B0.0%2026-09-10 evaluatedindependent evaluatorsource
gdpval_aaLFM2.5-2.6B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondLFM2.5-2.6B55.76%2026-09-10 evaluatedindependent evaluatorsource
scicodeLFM2.5-2.6B14.35%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