Source-first local LLM release notes, runtime updates, reproducible benchmarks, and practical comparisons.https://localllmupdates.comenhttps://localllmupdates.com/updates/#ollama-v0-33-2-desktop-maintenancehttps://localllmupdates.com/updates/#ollama-v0-33-2-desktop-maintenanceOllama restores following the system appearance, hands the macOS app off to an already-running instance, and prevents model-catalog updates from interrupting in-flight Claude Desktop proxy requests.Thu, 27 Aug 2026 00:00:00 GMTollamamacosclaude-desktophttps://localllmupdates.com/updates/#ollama-v0-33-0-claude-desktop-caching-packaginghttps://localllmupdates.com/updates/#ollama-v0-33-0-claude-desktop-caching-packagingOllama's v0.33.0 release lets Claude Desktop use Ollama as a third-party gateway provider, fixes prompt caching for recurrent-layer models, and fixes Linux and Windows packaging.Fri, 21 Aug 2026 22:52:46 GMTollamahttps://localllmupdates.com/updates/#ollama-v0-32-15-metadata-qwenhttps://localllmupdates.com/updates/#ollama-v0-32-15-metadata-qwenOllama reports that v0.32.15 caches resolved model metadata, fixes chat and generate wedging after a parser error, and normalizes Qwen 3.8 system messages.Wed, 19 Aug 2026 17:25:16 GMTollamaqwen-3-8https://localllmupdates.com/updates/#ollama-v0-32-14-webp-qwenhttps://localllmupdates.com/updates/#ollama-v0-32-14-webp-qwenThe official notes list WebP transcoding for llama-server and Qwen renderer tolerance for non-leading system messages.Sat, 15 Aug 2026 19:41:23 GMTollamahttps://localllmupdates.com/updates/#ollama-v0-32-12-adds-qwen-3-8-27b-supporthttps://localllmupdates.com/updates/#ollama-v0-32-12-adds-qwen-3-8-27b-supportThe release adds Qwen3.8-27B support, with Ollama stating particular optimizations for Apple Silicon.Fri, 14 Aug 2026 16:37:00 GMTqwen-3-8-27bollamahttps://localllmupdates.com/updates/#ollama-v0-32-10-nvfp4-prefillhttps://localllmupdates.com/updates/#ollama-v0-32-10-nvfp4-prefillOllama reports about 7–8% faster prefill on the Qwen3.6 and Muse Glimmer NVFP4 MLX models with a global scale, changes the default to 1.0 instead of 1.1 for models without `repeat_penalty`, and fixes a shared-digest blob-verification case.Wed, 12 Aug 2026 22:36:49 GMTollamahttps://localllmupdates.com/updates/#ollama-v0-32-9-adds-the-nemotron-3-architecturehttps://localllmupdates.com/updates/#ollama-v0-32-9-adds-the-nemotron-3-architectureOllama's release notes add the Nemotron 3 architecture and describe NVIDIA Nemotron 3.5 Lightning as an open 30B MoE model with 3B active parameters.Tue, 11 Aug 2026 13:23:33 GMTnemotron-3-5-lightningollama