Ollama v0.34.2 fixes MLX speculative decoding memory growth and adds first-run setup | Tool Update
Ollama published release v0.34.2 for the local open model inference runtime on September 15, 2026 (21:21, timestamp timezone inferred as UTC from GitHub's rele…
Published on MyPrivateClaw
Sep 18, 2026, 6:13 AM UTC
Coverage date
Sep 17, 2026
Last updated
Sep 18, 2026, 6:13 AM UTC
News summary
Ollama published the Ollama release v0.34.2 for the local open model inference runtime on September 15, 2026 (21:21 UTC, with the exact timezone inferred as UTC from GitHub's release display). According to the release notes, v0.34.2 fixed excessive memory growth during long generations with MLX speculative decoding, and the release also updated llama.cpp. The same release added first run setup when running ollama, with options to sign in or continue locally, and it shares setup completion with the desktop app on macOS and Windows. Why it matters The v0.34.2 changes operate at the runtime layer rather than the model layer. The fix for excessive memory growth during long generations with MLX speculative decoding addresses memory behavior in a generation path rather than model output or measured model quality, which the evidence packet does not claim to improve or degrade. The added first…