v2.0.0 v0.2.85
mlc-ai/web-llmSep 4, 2026 → Sep 8, 20264 days between releases
What's New
Updates TVM runtime and OPFS handles; incompatible with v2.0.0 server SDK
New Features
- OPFS synchronous access handles and committed cache records
- OpenAI-compatible `created` timestamps in seconds
- Service-worker handler registration for restarted workers
- Worker model state preservation to avoid reloads
Improvements
- Updated bundled TVM web runtime to `0.27.0-dev0`
- Improved stability by preventing hangs and rejecting invalid parameters
Bug Fixes
- Prevent hangs when initializing invalid structural-tag constraints
- Reject NaN generation parameters
- Render example output safely and correct cached-model `appConfig` examples
Upgrade Notes
Major version mismatch: v2.0.0 targets server architecture (MCP/Skybridge) while v0.2.85 targets client-side runtime (TVM/OPFS). These appear to be different components with no direct upgrade path.
The Two Releases
v2.0.0: North Face
Sep 4, 2026· 2 months ago
A major version release rebuilding the server architecture to support the new MCP 2026-07-28 revision. It introduces a new `Skybridge` app structure, a `setup` function for initialization, and a new testing library.
Full release notesv0.2.85
Sep 8, 2026· 1 month ago
This release updates the bundled TVM runtime and introduces synchronous OPFS access handles. It improves stability by preventing hangs and rejecting invalid parameters, while ensuring OpenAI-compatible API timestamps and preserving worker model state.
Full release notes