Unsloth adds AMD support for LLM training and claims 2x faster runs with 70% less VRAM
danielhanchen · x · 2026-07-21
Unsloth says it now supports training and running LLMs on AMD hardware.
- Collaborated with AMD to enable 500+ models on AMD GPUs
- Works on Windows, WSL, and Linux
- Claims up to 2× faster training with 70% less VRAM, with no accuracy loss
- Supports Radeon, Instinct, Ryzen, and data center GPUs
- Also adds optimized ROCm builds for GGUF and Safetensors inference
The post also highlights Unsloth’s local training/inference stack, including tool-call healing, code execution, secure web search, remote APIs, and HTTPS deployment.
Related event: Unsloth Expands AMD Support for LLM Training and Inference(3 posts)→
More from coding & agent
- OpenAI ships GPT-6.1 Terra; Brockman says its computer use now beats his own speed — gdb · 2026-10-27
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27