{"version":1,"type":"story","url":"https://digestai.news/story/atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b","json":"https://digestai.news/story/atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b.json","markdown":"https://digestai.news/story/atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b.md","slug":"atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b","headline":"AtomicChat releases quantized GGUF files for Nanbeige4.3-33B","summary":"AtomicChat has released quantized GGUF versions of Nanbeige's 33B dense agent and reasoning model, Nanbeige4.3-33B. Built on release day, 25 October, these files are derived from the original BF16 weights and are available under the Apache-2.0 license. The model features a Llama architecture with a 512K context window and has thinking enabled by default.\n\nThe release includes both standard llama.cpp presets (Q80, Q6K, Q5KM, Q4KM) and custom \"Atomic Dynamic\" (AD) files that use a specific importance matrix to preserve quality. According to AtomicChat's measurements, Nanbeige4.3-33B is unusually sensitive to quantization; standard Q6K files only match the original model's next-token predictions 95.5% of the time, whereas the AD-Q6K-Q5K file reduces KL divergence by 39% compared to the standard Q6K.\n\nUsers can run these models via llama-server or Atomic Chat. The article notes that files below 4 bits (such as AD-IQ3XXS) suffer significant performance drops, with top-1 accuracy falling below 85%. For optimal results, AtomicChat recommends specific sampling parameters: temperature 0.6 and top-p 0.95 for reasoning and chat, or temperature 1.0 and top-p 0.95 for agentic tasks.","keyPoints":["AtomicChat released quantized GGUF files for Nanbeige4.3-33B on 25 October.","Custom AD files show significantly lower KL divergence than standard llama.cpp presets.","The model is sensitive to quantization, with standard Q6K achieving only 95.5% top-1 accuracy."],"whyItMatters":"These quantized files allow developers to run a high-performance 33B reasoning model on local hardware with limited VRAM, offering a practical alternative to cloud-based inference for agentic applications.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Nanbeige","AtomicChat"],"models":["Nanbeige4.3-33B","Qwen3.8-27B"],"people":[]},"firstPublishedAt":"2026-10-25T21:33:29Z","updatedAt":"2026-10-25T21:33:29Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"huggingface.co","title":"Nanbeige4.3-33B Quantised GGUF","url":"https://huggingface.co/AtomicChat/Nanbeige4.3-33B-GGUF","publishedAt":"2026-10-25T21:33:29Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xfeho8/nanbeige4333b_quantised_gguf/","points":null}],"thread":{"title":"AtomicChat's Quantized Model Rollout Saga","url":"https://digestai.news/thread/atomicchat-releases-quantized-ling-3-1-flash-models-with-8688-token-accuracy","storyCount":2},"cite":{"text":"Digest AI, \"AtomicChat releases quantized GGUF files for Nanbeige4.3-33B\", 25 October 2026, https://digestai.news/story/atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b","publisher":"Digest AI","title":"AtomicChat releases quantized GGUF files for Nanbeige4.3-33B","datePublished":"2026-10-25T21:33:29Z","url":"https://digestai.news/story/atomicchat-releases-quantized-gguf-files-for-nanbeige4-3-33b"},"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"}