v0.18.3

Ollama Releases / 3/26/2026

📰 NewsDeveloper Stack & InfrastructureTools & Practical Usage

共有:

Key Points

Ollama v0.18.3 adds a new capability to launch models directly from VS Code using the `ollama launch vscode` command.
The release includes documentation updates for Claude Code/Openclaw to support web search and non-interactive operation.
It optimizes the launch command by skipping redundant configuration writes when the selected model hasn’t changed.
It improves ML performance by sharing the KV cache across conversations that share common prefixes (via the `mlxrunner` component).
The desktop app is fixed to avoid getting stuck loading when `OLLAMA_HOST` is set to an unspecified bind address.

Ollama can now directly launch models in VS Code.

ollama launch vscode # or code

docs: update claude code and openclaw for web search and non-interactive mode by @ParthSareen in #14922
cmd/launch: skip redundant config writes when model unchanged by @hoyyeva in #14941
mlxrunner: share KV cache across conversations with common prefixes by @jessegross in #14887
app: fix desktop app stuck loading when OLLAMA_HOST is an unspecified bind address by @hoyyeva in #14885
parsers: robust xml tool repair by @BruceMacD in #14961
add ability to turn on debug request logging by @drifkin in #14106
Fix mlxrunner subprocess deadlocks by @jessegross in #14919
launch: skip openclaw gateway health check when no daemon install by @BruceMacD in #14984
docs: nemoclaw integration by @BruceMacD in #14962
mlx: update as of 3/23 by @dhiltgen in #14789
docs: add Claude Code with Telegram guide by @ParthSareen in #15026
mlxrunner: support partial match on pure transformer caches by @jessegross in #14985
mlx: add mxfp4/mxfp8/nvfp4 importing by @pdevine in #15015
integration: improve ability to test individual models by @dhiltgen in #14948
ci: fix windows cgo compiler error by @dhiltgen in #15046

Full Changelog: v0.18.2...v0.18.3

Dev.to

Dev.to

Dev.to

Dev.to

Dev.to