Unsloth AI

We're releasing our final update to Qwen3.5 GGUFs for improved performance. - Qwen3.5 GGUFs now use our new iMatrix data for better chat, coding & tool use. - New improved quant algorithm - Re-download 35B, 27B, 122B GGUFs: https://lnkd.in/g3wWhYaF Guide: https://lnkd.in/gKdmXbDk

Qwen3.5-35B-A3B-GGUF is now the #3 trending model on Hugging Face! 💜 Run the LLM locally on 22GB RAM: https://lnkd.in/g5gchScH Read our guide: https://lnkd.in/gKdmXbDk

Qwen3-Coder-Next GGUF is now the most downloaded model on Unsloth! 🚀 The 80B coding LLM runs on a 36GB RAM Mac / device. Use via Claude Code and Codex locally. Qwen 3-Coder-Next GGUF on Hugging Face: https://lnkd.in/gcH_pGxa To learn how to run the model and see quant benchmarks (4-bit, 3-bit, etc.), read our guide: https://lnkd.in/gtTdiKT4

You can finetune AI models with Unsloth AI + Hugging Face jobs, and right now it’s FREE! We're giving away GPU credits on HF Jobs + Unsloth. Train a 1.2B model like Liquid AI LFM2.5-1.2B with one command. Ship it to your phone, laptop, or API. No infra. No setup. Just prompt your coding agent and go. Sign up here https://lnkd.in/egrQxvwv

100,000+ models trained with Unsloth have been open-sourced on Hugging Face! 🤗 Run these: 1. TeichAI GLM-4.7-Flash distilled from Claude 4.5 Opus (high) 2. Zed - Qwen Coder 7B fine-tuned for stronger coding 3. DavidAU - Llama-3.3-8B distilled from Claude 4.5 Opus (high) 4. huihui - gpt-oss made “abliberated” See all the 100K latest models fine-tuned with Unsloth here: https://lnkd.in/g8CZZf9W

You can now run MiniMax-2.5 locally! 🚀 At 230B parameters, it's the strongest LLM under 700B params. Run on a 128GB Mac for 20 tokens/s via Dynamic 3/4-bit precision. Run at 8-bit on 256GB RAM/VRAM. The model delivers SOTA in agentic coding & chat performance for open models. Guide: https://lnkd.in/gj_pwtAp GGUF: https://lnkd.in/gBbgJJps

You can now run GLM-5 locally! 🔥 GLM-5 is a new open SOTA agentic coding & chat LLM with 200K context. Runs on a 256GB Mac or RAM/VRAM setups. We shrank the 744B model from 1.65TB to 241GB (-85%) via Dynamic 2-bit. Guide: https://lnkd.in/gFaD9WCW GGUFs to run: https://lnkd.in/gueeThhQ

You can now train MoE models 12× faster with 35% less VRAM via our new Triton kernels! 🤗 We collabed with Hugging Face to enable you to train DeepSeek, Qwen3, GLM and gpt-oss all faster on Unsloth AI with no accuracy degradation. Train gpt-oss faster, on just 12.8GB VRAM with our free notebook. ⭐ Blog + notebooks: https://lnkd.in/dzNcZhDS GitHub: https://lnkd.in/gyaDBTxK

DeepSeek releases DeepSeek-OCR 2. 🐋 The new 3B model achieves SOTA visual, document and OCR understanding. DeepEncoder V2 is introduced which enables the model scan images in same logical order as humans, boosting OCR accuracy. Instead of traditional vision LLMs which read an image in a fixed grid (top-left → bottom-right), DeepEncoder V2 first builds a global understanding, then learns a human-like reading order - what to attend to first, next, and so on. This improves OCR on complex layouts helping it follow columns, link labels to values, read tables coherently, and handle mixed text + structure more reliably. DeepSeek AI-OCR 2 outperforms Gemini 3 Pro on benchmarks and is >4% improvement over the previous DeepSeek-OCR. You can now run and fine-tune DeepSeek-OCR 2 with Unsloth AI and our guide. Guide + Notebooks: https://lnkd.in/giMNuyuQ Model: https://lnkd.in/gRszxcye