LFM2.5 VL 1.6B
Jan 2026
LFM2.5 VL 1.6BCompare inon
LFM2.5 VL 1.6B measured on 8 of 104 Chat & Knowledge benchmarks8
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Legacy · saturated or no longer reported1
No benchmark matches
—per task?
self-host onlyprice
—output speed?
33Kcontext
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGPT-6 Lunalow×4—$0.005Function calling only with none · Chat Completions+3IOpenK2 Horizon 7Bthinking×3.6—$0.30 / $1.20Memory (4-bit): 0.8 GB → 4.5 GB+1BetterGPT-5.6 Lunalow×3.9—$0.01BetterGPT OSS 20BhighOpen×2.2—$0.011Memory (4-bit): 0.8 GB → 13 GB+1
BetterMistral Small 4thinkingOpen+94%—$0.015Memory (4-bit): 0.8 GB → 72 GBModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
GPT-6 Lunalow×4—$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow×3.9—$0.01GPT OSS 20BhighOpen×2.2—$0.011Memory (4-bit): 0.8 GB → 13 GB+1Mistral Small 4thinkingOpen+94%—$0.015Memory (4-bit): 0.8 GB → 72 GB
Nemotron 3 Nano 30B A3BthinkingOpen×2.2—$0.017Memory (4-bit): 0.8 GB → 19 GB+1ModelScoreCostPer taskWhat changes
IK2 Horizon 7BthinkingOpen×3.6—$0.30 / $1.20Memory (4-bit): 0.8 GB → 4.5 GB+1
Gemma 4 12BthinkingOpen×3.1—$0.10 / $0.30Memory (4-bit): 0.8 GB → 6 GB
Qwen 3.5 9BthinkingOpen×2.9—$0.21Memory (4-bit): 0.8 GB → 4.8 GBIK2 Horizon 3.7BthinkingOpen×2.7—$0.30 / $1.20Memory (4-bit): 0.8 GB → 2.5 GB+1TFalcon H1R 7BthinkingOpen×2.2——Memory (4-bit): 0.8 GB → 3.8 GB+1ModelScoreCostPer taskWhat changes
LFM2.5 8B A1BthinkingOpen+35%——Memory (4-bit): 0.8 GB → 4.2 GB+1LFM2.5 1.2B Instructnon-thinkingOpen+32%——No image inputLFM2.5 1.2B ThinkingthinkingOpen+22%——No image inputLFM2 1.2BOpensame——No image inputLFM2 2.6BOpensame——Memory (4-bit): 0.8 GB → 1.3 GB+1LFM2 8B A1BOpensame——Memory (4-bit): 0.8 GB → 4.2 GB+1LFM2 24B A2Bnon-thinkingOpensame—$0.03 / $0.12Memory (4-bit): 0.8 GB → 14 GB+1All 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
in Chat & KnowledgeAA Intelligence Index 4.3.2≈4.8#327
AA Omniscience Index-84.4#321
AA Omniscience · Accuracy5.8#312
Humanity's Last Exam5.1#269
GPQA Diamond28.9#302
IFBench (instructions)33.1#197
CritPt (physics)0#180
FAQ
How much does LFM2.5 VL 1.6B cost?›
There is no per-unit API price: self-host only.
Is LFM2.5 VL 1.6B free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is LFM2.5 VL 1.6B, and what is it best at?›
#267 of 334 in Chat & Knowledge, #210 of 239 in Coding, #227 of 242 in Agents & Tool Use, #195 of 245 in Writing & Language, #122 of 124 in Vision & Documents.
Is LFM2.5 VL 1.6B better than Llama 3.1 Nemotron Nano 4B v1.1?›
On Humanity's Last Exam, LFM2.5 VL 1.6B scores 5.1 vs 5.2 — about the same.
What is the context window of LFM2.5 VL 1.6B?›
33K tokens.
When was LFM2.5 VL 1.6B released? Is there a newer version?›
Released 2026-01-05. It is the current version in its line.
Is LFM2.5 VL 1.6B open source?›
Yes, the weights are open (lfm1.0 licence).
Can I run LFM2.5 VL 1.6B locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can LFM2.5 VL 1.6B read images?›
It accepts text and images as input.
What is LFM2.5 VL 1.6B?›
LFM2.5 VL 1.6B is an open-weights language model from Liquid AI, released Jan 2026.
What are the best alternatives to LFM2.5 VL 1.6B?›
GPT-6 Luna (better), K2 Horizon 7B (open weights).