openclaw v2026.6.10-beta.2 release notes: what's new & changelog

openclaw v2026.6.10‑beta.2 delivers automatic fast mode for short conversational turns, tighter model routing, and a suite of reliability and safety fixes.

Key Changes

• Automatic fast mode for talks – Short turns now auto‑enable fast mode and revert to normal mode after a configurable cutoff (default 60 s). Fast mode state survives retries, fallbacks, and progress events. (#85104)
• More reliable model routing – Zai model synthesis, GLM overload failover, and native reasoning‑level selection now consistently follow the active model catalog. (#94461, #93241, #94067, #94136)
• Safer session and channel state – Channel switches reset stale origin fields; cron delivery awareness stays attached to the target session. (#95328, #93580)
• Trusted policies survive hook composition – Hook registries now retain required trusted‑policy entries for approval‑sensitive flows. (#94545)
• Model catalog improvements – Correct baseUrl fallback for synthesized GLM‑5 models, exposure of GLM‑5.2 reasoning levels, and runtime catalog usage for native /think menus. (#94461, #94136, #94067)
• Various fixes – Bounded fallback cutoffs, visible progress events, correct Zhipu GLM overload detection, and reset of stale per‑channel origin fields on channel switches. (#85104, #94545, #93241, #95328)

Impact

• User experience – Conversations with short exchanges become faster and more responsive thanks to auto fast mode, while longer sessions retain normal performance.
• Model reliability – Overload situations for Zhipu GLM and GLM‑5 models now trigger proper failover, reducing failed completions.
• Channel consistency – Switching between channels (e.g., Slack → Telegram) no longer leaks stale channel identifiers, fixing context‑aware features such as reactions and threading.
• Policy enforcement – Composed hook registries keep trusted policies intact, preserving security guarantees for approval‑sensitive workflows.
• Developer ergonomics – Native /think menus now reflect the full reasoning level set of live‑discovered models, and GLM‑5.2 users can select granular reasoning levels (off, low, high, max).

FAQ

What's new in v2026.6.10-beta.2? This release adds automatic fast mode for short talks, improves model routing and overload handling, makes session and channel state safer, and fixes several edge‑case bugs around fast mode, model catalogs, cron delivery awareness, and channel origin fields.

Are there any breaking changes? No breaking changes are introduced in this beta release; all changes are additive or bug‑fixes and remain backward compatible.

How do I upgrade? Upgrade by installing the new npm package (npm install openclaw@2026.6.10-beta.2) or pulling the latest release tarball; no migration steps or configuration changes are required.