Tools
Ollama 0.34.1 graduates MLX safetensors creation and cuts /api/tags cold time from 3.1 s to 294 ms
Released 2026-09-14, v0.34.1 takes MLX safetensors `ollama create` out of experimental, while GGUF model creation now requires llama.cpp tooling for safetensor conversion and quantization. /api/tags is roughly 10x faster on large model libraries (3.1 s to 294 ms cold in their testing) and model capabilities are reported consistently. Runaway repeat-token detection now needs 100 repeated tokens before firing, cutting false positives on OCR-style output, and typical_p is deprecated for new models while existing GGUF models keep it.
Source
↳ Follow the thread