[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-15b-african-language-model-beats-google-meta-and-alibaba":10,"sections":44},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":34,"tags":35,"sources":39,"feedback":43,"feedback_at":22,"cost_usd":43,"total_tokens":43},7507,"15b-african-language-model-beats-google-meta-and-alibaba","1.5B African Language Model Beats Google, Meta and Alibaba","Vambo AI's 1.5B parameter MORENA model, built from scratch for 12 African languages, beats larger Google, Meta and Alibaba models on key benchmarks.","A 1.5 billion parameter model built for African languages just beat systems from Google, Meta and Alibaba that are up to eight times its size.\n\nVambo AI released MORENA, a 1.5B model covering 12 African languages plus English, French and code. Rather than fine-tuning an existing Llama or Gemma checkpoint, the team trained it from scratch with a custom tokenizer built around African languages instead of one designed for English and programming text. Vambo says MORENA scored 1.408 bits-per-byte on a key benchmark, the best of 26 models tested, edging out its closest rival's 1.423 despite that model carrying more than five times as many parameters. The company also released smaller 0.5B and 0.2B versions, the latter aimed at tasks like keyboard input and speech-recognition rescoring.\n\nMost African-language AI projects bolt extra training onto an existing Llama or Gemma checkpoint, inheriting a vocabulary built for English and code that wastes tokens on African text. MORENA's custom tokenizer needs up to 1.53 times fewer tokens than Llama 3.2 for the same passages, which means cheaper training and inference without depending on a foreign base model's assumptions in the first place. That design choice - not just the benchmark scores - is the more interesting bet here.\n\nThe whole project reportedly ran to about 22,000 GPU hours and $40,000 in compute, with support from UNDP, AIHub4SD and CINECA - a reminder that closing Big Tech's language gaps doesn't always require Big Tech's budget.","[\"ai\",\"african-languages\",\"language-models\",\"startups\"]","2026-09-23T21:25:00.000Z","2026-09-23T23:05:04.444Z","2026-09-23T23:05:08.402Z","published",null,[24,30],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"The dek claims MORENA beats bigger rivals from Google, Meta, and Alibaba, but the body never mentions Alibaba or any Alibaba model comparison at all — either add a supporting Alibaba benchmark detail or drop Alibaba from the dek.","resolved",{"id":31,"reviewer":26,"round":32,"reason":33,"status":29},"editor-r2",2,"The body now names Alibaba but immediately undercuts the claim with 'though it does not specify which of their systems it tested against' — this reads as an admission of incomplete reporting, so either drop that hedge and cite Vambo's claim plainly or remove the Google\u002FMeta\u002FAlibaba comparison from the headline and dek since it isn't backed by specific benchmark data.","ai",[34,36,37,38],"african-languages","language-models","startups",[40],{"name":41,"url":42},"TechRadar","https:\u002F\u002Fwww.techradar.com\u002Fpro\u002Fafricas-ai-moment-has-arrived-this-1-5b-model-built-from-scratch-for-12-african-languages-beats-google-meta-and-alibaba-while-being-8x-smaller",0,{"sections":45},[46,50,55,60,65,70,75,80,85,90,94,99,104,109],{"name":47,"slug":34,"count":48,"latest_published_at":49},"AI",4363,"2026-09-23T23:03:26.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Security","security",718,"2026-09-23T20:24:22.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Policy","policy",380,"2026-09-23T22:53:43.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Deals","deals",220,"2026-09-23T22:00:00.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Hardware","hardware",172,"2026-09-23T20:44:59.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",134,"2026-09-23T09:00:00.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Consumer Tech","consumer-tech",111,"2026-09-23T17:04:48.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Software","software",85,"2026-09-23T20:00:00.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Dev Tools","dev-tools",79,"2026-09-22T22:21:13.000Z",{"name":91,"slug":38,"count":92,"latest_published_at":93},"Startups",66,"2026-09-23T17:28:38.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Gaming","gaming",45,"2026-09-22T15:35:06.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"General","general",43,"2026-09-21T23:48:56.000Z",{"name":105,"slug":106,"count":107,"latest_published_at":108},"Reviews","reviews",27,"2026-09-22T13:00:00.000Z",{"name":110,"slug":111,"count":112,"latest_published_at":113},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]