Fine-tuned 350M browser-agent model (LFM2.5-350M). 91.2% action accuracy on Mind2Web, runs in-browser via wllama WebAssembly.
wasm text-generation lora mlx fine-tuning on-device lfm web-agent apple-silicon browser-agent gguf unsloth wllama lfm2 mind2web mlx-tune
-
Updated
Jul 8, 2026 - Python