Source: github.com/cline/cline/releases/tag/sdk%2Fsdk%2Fv0.0.95
Requests sent through OpenAI-compatible endpoints to GPT-6 model IDs, including gpt-6-sol, gpt-6.1-sol, gpt-6-1-sol and later gpt-5 through gpt-9 IDs, now use max_completion_tokens. Azure's gpt-35-turbo keeps max_tokens.
The hub daemon now records which client started it through CLINE_HUB_LAUNCHER_CLIENT, CLINE_HUB_LAUNCHER_VERSION and CLINE_HUB_LAUNCHER_TRANSPORT, and publishes these as OpenTelemetry resource attributes. SSH remote hubs report the local client with transport ssh. The @cline/shared package adds withHubLauncherIdentity, claimHubLauncherIdentity and readHubLauncherIdentity, and the telemetry self-check report now includes traces_exporter.
The model catalog was refreshed from 6,952 to 7,041 models, and the CommandCode and Reflection providers were added. The cline provider's bundled default moves back to Claude Sonnet 5.5 from Gemini 4 Argon, and default models changed for several other providers. OpenCode Go's base URL is now pinned at https://opencode.ai/inference/go/v1 because models.dev moved its provider-level URL to an OpenAI-only path.
Why it matters if you build agents
If your agents call GPT-6 models through OpenAI-compatible endpoints, requests to OpenAI and Azure should no longer fail on the token parameter.
If you rely on default models, check which model the cline provider or the affected providers now select, since several defaults changed in this release.
Pinning the OpenCode Go base URL means that setting will not follow future changes to models.dev.
Get stories like this every weekday
One short email with the day's agent tooling news.