{"version":1,"type":"story","url":"https://digestai.news/story/nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents","json":"https://digestai.news/story/nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents.json","markdown":"https://digestai.news/story/nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents.md","slug":"nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents","headline":"Nanbeige releases Nanbeige4.3-33B teacher model for lightweight agents","summary":"Nanbeige announced the Nanbeige4.3-33B model, a high‑capacity teacher designed to seed a pipeline that distills agentic abilities into smaller models such as a forthcoming 3B version. The 33B‑parameter model aims to set a new performance bar among open models in the ~30B parameter class, excelling at complex coding and professional cowork tasks.\n\nThe team evaluated Nanbeige4.3-33B against frontier models Qwen3.8-27B, Muse‑Glimmer‑30B and Gemma4‑31B using a suite of benchmarks that include Terminal‑Bench, DeepSWE, SWE‑Atlas‑QnA, NL2Repo, Automation Bench v1.0.6 (600 public tasks), JobBench (35 occupations), GDPval, Claw‑Eval General (157 tasks), Toolathlon‑Verified, YC‑bench (10 recent turns) and others. The model led on most tests and showed strong multi‑turn tool‑use performance when the “preservethinking” flag is enabled. It supports a context window of up to 524,288 tokens and can be run with Transformers, SGLang or vLLM.","keyPoints":["Nanbeige4.3-33B is a 33B‑parameter teacher model targeting lightweight agent distillation.","It outperforms Qwen3.8-27B, Muse‑Glimmer‑30B and Gemma4‑31B on most benchmarks, including 35 occupations in JobBench.","The model supports up to 524,288 tokens and can be served via Transformers, SGLang or vLLM."],"whyItMatters":"A strong open‑weight 33B teacher model gives researchers and developers a new baseline for building cost‑effective, high‑performing agentic systems without relying on proprietary frontier models.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Nanbeige","Hugging Face"],"models":["Nanbeige4.3-33B","Qwen3.8-27B","Muse-Glimmer-30B","Gemma4-31B"],"people":["Nanbeige Team"]},"firstPublishedAt":"2026-10-25T12:01:34Z","updatedAt":"2026-10-25T12:01:34Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"huggingface.co","title":"Nanbeige/Nanbeige4.3-33B · Hugging Face","url":"https://huggingface.co/Nanbeige/Nanbeige4.3-33B","publishedAt":"2026-10-25T12:01:34Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xf0kz4/nanbeigenanbeige4333b_hugging_face/","points":null}],"thread":null,"cite":{"text":"Digest AI, \"Nanbeige releases Nanbeige4.3-33B teacher model for lightweight agents\", 25 October 2026, https://digestai.news/story/nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents","publisher":"Digest AI","title":"Nanbeige releases Nanbeige4.3-33B teacher model for lightweight agents","datePublished":"2026-10-25T12:01:34Z","url":"https://digestai.news/story/nanbeige-releases-nanbeige4-3-33b-teacher-model-for-lightweight-agents"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}