LFM2 2.6B
Sep 2025
LFM2 2.6BCompare inon
Benchmarks of LFM2 2.6B12
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Coding1
Legacy · saturated or no longer reported4
No benchmark matches
—per task?
self-host onlyprice
—output speed?
33Kcontext
Ranking?
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGranite 4.2 3BthinkingOpen×2.4—$0.006Memory (4-bit): 1.3 GB → 1.8 GBOOpenMiniCPM5 1Bnon-thinking×69.8——
BetterGemma 4 E4BthinkingOpen×2—$0.008Memory (4-bit): 1.3 GB → 4 GB
BetterMiMo v2.5thinkingOpen+97%—$0.019Memory (4-bit): 1.3 GB → 186 GBBetterGranite 4.2 8BthinkingOpen+97%—$0.024Memory (4-bit): 1.3 GB → 4.4 GBModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
Granite 4.2 3BthinkingOpen×2.4—$0.006Memory (4-bit): 1.3 GB → 1.8 GBGemma 4 E4BthinkingOpen×2—$0.008Memory (4-bit): 1.3 GB → 4 GBMiMo v2.5thinkingOpen+97%—$0.019Memory (4-bit): 1.3 GB → 186 GBGranite 4.2 8BthinkingOpen+97%—$0.024Memory (4-bit): 1.3 GB → 4.4 GB
Claude Haiku 5.5low+40%—$0.024Remove sampling parameters+5ModelScoreCostPer taskWhat changes
OMiniCPM5 1Bnon-thinkingOpen×69.8——AG9v3 3BOpen×5.4——Memory (4-bit): 1.3 GB → 1.5 GBlLFM2.5 2.6BthinkingOpen×3.9——OMiniCPM5 2BthinkingOpen×2.9——Granite 4.2 3BthinkingOpen×2.4—$0.006Memory (4-bit): 1.3 GB → 1.8 GBModelScoreCostPer taskWhat changes
LFM2.5 VL 1.6Bnon-thinkingOpen-34%——LFM2.5 1.2B ThinkingthinkingOpen-33%——LFM2 8B A1BOpen-32%——Memory (4-bit): 1.3 GB → 4.2 GBLFM2.5 1.2B Instructnon-thinkingOpen-26%——LFM2 1.2BOpen-23%——LFM2 24B A2Bnon-thinkingOpen-9%—$0.03 / $0.12Memory (4-bit): 1.3 GB → 14 GBLFM2.5 8B A1BthinkingOpen+34%——Memory (4-bit): 1.3 GB → 4.2 GBAll Liquid AI models →Run it yourself
estimateFits on✓ 8 GB (phone, laptop) · full quality✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
Licencelfm1.0commercial use with conditions
Benchmarks
all categoriesChat & Knowledge
Humanity's Last Exam5.5#261
AA Omniscience Index-54.1#232
CritPt (physics)0#180
AA Intelligence Index 4.3.2≈5.2#316
AA Omniscience · Accuracy5.4#313
GPQA Diamond30.6#297
AIME 2025 · AA archived8.3#128
Coding
LiveCodeBench · AA archived8.1#147
Agents & Tool Use
τ²-bench · AA13.5#214
Writing & Language
IFBench (instructions)19.5#245
FAQ
How much does LFM2 2.6B cost?›
There is no per-unit API price: self-host only.
Is LFM2 2.6B free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is LFM2 2.6B, and what is it best at?›
#259 of 334 in Chat & Knowledge, #203 of 239 in Coding, #211 of 242 in Agents & Tool Use, #242 of 245 in Writing & Language.
Is LFM2 2.6B better than Claude Opus 4.6?›
On AA Omniscience · Hallucination Rate, LFM2 2.6B scores 62.9 vs 62.8 — about the same.
What is the context window of LFM2 2.6B?›
33K tokens.
When was LFM2 2.6B released? Is there a newer version?›
Released 2025-09-23. It is the current version in its line.
Is LFM2 2.6B open source?›
Yes, the weights are open (lfm1.0 licence).
Can I run LFM2 2.6B locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can LFM2 2.6B read images?›
No, it accepts text only.
What is LFM2 2.6B?›
LFM2 2.6B is an open-weights language model from Liquid AI, released Sep 2025.
What are the best alternatives to LFM2 2.6B?›
Granite 4.2 3B (better), MiniCPM5 1B (open weights).