ExoProduct update
Exo · v1.0.70
EXO v1.0.70 supports vision models, using the vision processors in Qwen3.5, Kimi K2.5 and Gemma 4, sends PDFs as both text and image, and adds support for Gemma 4, Minimax M2.7 and Qwen3.6. Flash Attention is implemented for Qwen3.5 and Gemma 4, which the release notes say cuts peak memory consumption by 3-6x, and memory leaks in the Rotating and Arrays caches are fixed.
Saved . A shortened excerpt from GitHub Releases; the original publisher retains authorship.