Inkampani esafufusa yobuhlakani bokwenziwa i-Liquid AI ikhiphe izolo ngesikhathi sendawo imodeli ye-ejenti evulekile i-LFM2.5-2.6B. Njengoba inamapharamitha ayizigidi eziyizinkulungwane ezingu-2.6 kuphela, i-LFM2.5-2.6B ifaneleka kakhulu ukufakwa nokusebenza endaweni kudivayisi, futhi ingasebenza ngisho nakuma-smartphone; ngesikhathi esifanayo inamandla anele ukusekela ukuhlela, ukubiza amathuluzi kanye nokugeleza komsebenzi kwama-ejenti okucubungula izinyathelo eziningi.
I-IT之家 ithole ukuthi i-LFM2.5-2.6B yaqeqeshwa kusengaphambili kusethi yedatha enamathokheni angu-34T, futhi yathuthukisa ukusekelwa kwezilimi ezisebenzisa izinhlamvu ezingezona ezesiLatini ngokwandisa i-tokenizer ekhona. I-Liquid AI inikeza kanyekanye ku-platform i-Hugging Face izinguqulo eziyisisekelo kanye neziqeqeshwe ngemuva zale modeli.
Ngokwemininingwane enikezwe yi-Liquid AI, ngesikhathi i-LFM2.5-2.6B ihlolwa ngokwesisekelo kanye namamodeli amakhulu afana ne-gemma-4-E2B-it, i-gemma-4-E4B-it, i-Qwen3.5-4B kanye ne-Qwen3.5-9B, ihole kuzo zonke izigaba zokusebenzisa imiyalelo nakuzo cishe zonke izigaba zokusebenzisa amathuluzi, yahlulwa kuphela yi-Qwen3.5-9B ekuhlolweni kwe-BFCLv4.
I-LFM2.5-2.6B yadlula ngokuphelele womabili amamodeli e-Gemma emisebenzini yama-ejenti, futhi yenza ngokufana namamodeli e-Qwen. Emkhakheni we-STEM, yahola ekuhlolweni kwe-AA Omniscience, kodwa yahlulwa yi-Qwen3.5-9B kuphela ekuhlolweni kwezibalo. Ukuhlela kwaba ukuphela komkhakha lapho amamodeli amakhulu esaqhubeka nokuba nenzuzo.