{"rewrite":{"id":"r_657eaeaecb6033b6e33d993d","clusterId":"c_5a74e38f3d1da960ed73cb33","slug":"liquid-ai-releases-lfm2-5-models-with-top-japanese-performance","model":"deepseek-v4-flash","headline":"Liquid AI Releases LFM2.5 Models With Top Japanese Performance","summary":"Liquid AI has released the LFM2.5 family of on-device AI models, including a 1.2-billion-parameter Japanese language model and a 1.5-billion-parameter Japanese audio model. The company says the language model achieves top-tier performance among same-scale models on Japanese benchmarks, and the audio model handles speech recognition and dialogue without separate ASR or TTS components. The models are available as open weights on Hugging Face.","whyItMatters":"The release gives Japanese-market developers a compact, open-weight model that claims leading benchmark scores for its size, potentially accelerating on-device AI agents in automotive, manufacturing, and smart devices.","webCardHtml":"\u003cp\u003eLiquid AI, a startup spun out of MIT, announced the LFM2.5 family on Thursday. The lineup includes the LFM2.5-1.2B-JP-202606, a 1.2-billion-parameter language model, and the LFM2.5-Audio-1.5B-JP, a 1.5-billion-parameter audio model. Both are designed for edge deployment. The company stated the language model outperforms comparable models across Japanese knowledge, instruction following, mathematics, coding, and tool use. The audio model is end-to-end, eliminating separate speech recognition and text-to-speech components. The models are released as open weights under the LFM Open License v1.0 and are available on Hugging Face. Liquid AI is also working with AMD, Qualcomm, and Nexa AI to deploy the models on NPUs.\u003c/p\u003e","blueskyPost":"Liquid AI released the LFM2.5 family of on-device models, including a 1.2B Japanese language model and a 1.5B Japanese audio model. The company claims top-tier Japanese benchmark performance at their scale. Open weights on Hugging Face.","twitterPost":"Liquid AI released the LFM2.5 family of on-device models, including a 1.2B Japanese language model and a 1.5B Japanese audio model. The company claims top-tier Japanese benchmark performance at their scale. Open weights on Hugging Face.","threadsPost":null,"newsletterBlurb":"Liquid AI released the LFM2.5 family of on-device AI models, including a 1.2-billion-parameter Japanese language model and a 1.5-billion-parameter Japanese audio model. The company says the language model achieves top-tier performance among same-scale models on Japanese benchmarks. The models are available as open weights on Hugging Face.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/408/4408302/?rss\",\"title\":\"Liquid AI Releases Small Language Model and Audio Model with Top-Tier Japanese Performance\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4398,"outputTokens":646,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1780625898,"createdAt":"2026-06-05T02:11:54.000Z","publishedAt":"2026-06-05T02:16:15.000Z","updatedAt":"2026-06-05T02:16:15.000Z"},"cluster":{"id":"c_5a74e38f3d1da960ed73cb33","canonicalTitle":"Liquid AI、日本語で最高水準の性能を持つ小規模言語モデルと音声モデルを公開","representativeArticleId":"a_031864ff963a229e3c5824fa","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-06-05T00:31:19.000Z","lastSeenAt":"2026-06-05T00:31:19.000Z","updatedAt":"2026-06-05T02:16:15.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/408/4408302/?rss","title":"Liquid AI、日本語で最高水準の性能を持つ小規模言語モデルと音声モデルを公開"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["Liquid AI released the LFM2.5 family on Thursday, including a 1.2-billion-parameter Japanese language model and a 1.5-billion-parameter Japanese audio model.","The company says the language model outperforms comparable models on Japanese benchmarks for knowledge, instruction following, mathematics, coding, and tool use.","The audio model is end-to-end, handling speech recognition and dialogue without separate ASR or TTS components.","The models are released as open weights under the LFM Open License v1.0 and are available on Hugging Face.","Liquid AI is working with AMD, Qualcomm, and Nexa AI to deploy the models on NPUs."]}