[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-slideformer-lets-a-single-rtx-4090-fine-tune-123b-models":10,"sections":35},{"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":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},3161,"slideformer-lets-a-single-rtx-4090-fine-tune-123b-models","SlideFormer Lets a Single RTX 4090 Fine-Tune 123B Models","A new open-source system moves LLM fine-tuning off expensive server clusters and onto consumer GPUs by sliding computation across CPU and memory tiers.","A research team has published SlideFormer, a system that fine-tunes large language models exceeding 123 billion parameters on a single consumer GPU.\n\nSlideFormer treats the GPU as a sliding window, overlapping GPU computation with CPU updates and multi-tier I\u002FO through a lightweight asynchronous engine. A heterogeneous memory management scheme roughly halves peak CPU and GPU memory usage compared to baseline approaches. The system also ships optimized Triton kernels to clear throughput bottlenecks. Benchmarks show 1.40x to 6.27x higher throughput than baselines, and it supports batch sizes up to 8x larger and models up to 6x larger than what fits in GPU memory alone. Code is available on GitHub.\n\nFine-tuning has been the quiet tax on AI adoption — organizations that can afford to run a model often cannot afford to customize it without renting cloud infrastructure. SlideFormer pushes that threshold down to hardware a serious hobbyist or small lab might already own, which matters more than any marginal throughput number. Whether the gains hold on real-world training jobs with irregular data is something published benchmarks rarely tell you.\n\nSlideFormer joins a crowded field of memory-reduction techniques — quantization, LoRA, gradient checkpointing — but targets the system layer rather than the model layer, which is a meaningfully different bet on where the headroom actually lives.","[\"machine learning\",\"open-source\",\"hardware\",\"fine-tuning\"]","2026-07-01T04:00:00.000Z","2026-07-01T08:41:34.899Z","2026-07-01T08:41:37.876Z","published",null,[],"ai",[26,27,28,29],"machine learning","open-source","hardware","fine-tuning",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.16428",0,{"sections":36},[37,41,46,51,56,60,65,70,75,80,85,89,94,99],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":45},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":57,"slug":28,"count":58,"latest_published_at":59},"Hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":86,"slug":87,"count":83,"latest_published_at":88},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]