What's Changed

• x/create: quantize lm_head at 8-bit in the requested family by @jessegross in https://github.com/ollama/ollama/pull/17357
• test: harden flaky updater and transfer unit tests by @dhiltgen in https://github.com/ollama/ollama/pull/17378
• server: fix ps data race on scheduler loaded map by @dhiltgen in https://github.com/ollama/ollama/pull/17376
• qwen3_5: fix expert quantization handling and gather packed gate_up in one launch by @jessegross in https://github.com/ollama/ollama/pull/17336
• agent: permission skill loading by @ParthSareen in https://github.com/ollama/ollama/pull/17304
• cmd/tui: agent system prompt command by @ParthSareen in https://github.com/ollama/ollama/pull/17296
• mlx: keep loaded model memory resident by @dhiltgen in https://github.com/ollama/ollama/pull/17367
• x/create: quantize a draft model's output head at the requested type by @jessegross in https://github.com/ollama/ollama/pull/17383
• model: add Laguna MLX support by @dhiltgen in https://github.com/ollama/ollama/pull/17237

Full Changelog: https://github.com/ollama/ollama/compare/v0.32.3...v0.32.4-rc0