agents · review · first seen Oct 26, 09:32 UTC

Cline updates cloud task settings and model defaults

Cline v4.1.26 shipped separate cloud model selection, hidden cloud-session controls, a lower retry ceiling, and API behavior changes. Builders should check whether new cloud tasks and API-backed workflows still use the settings they intend.

L0 · primary1 source5 claims checked4 min read

What shipped

Cline v4.1.26 shipped a separate model setting for cloud tasks when cloud sessions are on. Settings > API Configuration shows a Local | Cloud toggle. The Cloud side selects the Cline model for new cloud tasks and saves that choice separately from local models. Clicking the model label in a cloud session’s composer opens the setting. Starting or resuming a cloud sandbox also shows a progress card with the repository, branch, and steps to start the sandbox, clone the repository, and start the agent. The card collapses once the agent starts. [1]

Cloud sessions now hide the Plan/Act toggle and auto-approve bar instead of displaying them as disabled controls. Their absence in a cloud session is the shipped interface behavior. The supplied excerpt does not describe a replacement control. [1]

Model requests now retry at most 3 times, rather than 5. The release notes also say that, when no model is chosen, the Cline provider again uses Claude Sonnet 5.5. That fallback is worth checking if you leave a model selection unset; it does not mean a model you explicitly selected has been replaced. [1][2]

GPT-6 models now work on OpenAI and Azure OpenAI-compatible endpoints. Separately, Gemini API keys no longer send native Google Search. These are distinct changes: one concerns endpoint compatibility, while the other removes a particular search path. Neither statement establishes what models or search alternatives are available in a specific account. [1][2]

Why it matters if you build

If you move between local work and cloud tasks, do not assume one model choice governs both. Check the Cloud side of API Configuration before starting a new task, and use the cloud composer’s model label to revisit that setting. The separate saved choices make an unset selection especially worth inspecting in light of the stated Cline-provider fallback. The source gives no price or quality comparison, so this is a configuration check, not a claim that one choice will cost less or work better. [1]

If you used the visible Plan/Act toggle or auto-approve bar to orient yourself, a cloud session will now look different. Do not treat the missing controls alone as evidence of a failed load. But do not infer an approval policy from the interface change either: the excerpt says the controls are hidden, not how cloud-session decisions are governed. Check that workflow before relying on a cue you used locally. [1]

A ceiling of 3 retries gives a failing model request fewer possible retries than the former ceiling of 5. The source does not specify retry delays, which failures qualify, or any billing effect. If you route GPT-6 through an OpenAI or Azure OpenAI-compatible endpoint, test your configured path. If a Gemini API-key task depended on native Google Search, test whether that task still does what you need; this release says that search path is no longer sent. [1][2][3]

How to try it today

1. With cloud sessions on, open Settings > API Configuration and find the Local | Cloud toggle.
2. Select the Cloud side and inspect the model chosen for new cloud tasks alongside your separate local choice.
3. Open a cloud session, click its composer model label, and check the model setting and the absence of the Plan/Act toggle and auto-approve bar.
4. Start or resume a cloud sandbox and observe its repository, branch, and progress steps on the card.
5. Where relevant, test your configured GPT-6 endpoint or a Gemini API-key task that previously depended on native Google Search. Check a failing model request against the new retry ceiling.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   Cline v4.1.26 adds a separate cloud model setting when cloud sessions are on.

   per GitHub releases · cline/cline (L0)
2. shipped L0 · primary

   Cloud sessions hide the Plan/Act toggle and auto-approve bar.

   per GitHub releases · cline/cline (L0)
3. shipped L0 · primary

   Model requests retry at most 3 times instead of 5.

   per GitHub releases · cline/cline (L0)
4. shipped L0 · primary

   GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints.

   per GitHub releases · cline/cline (L0)
5. shipped L0 · primary

   Gemini API keys no longer send native Google Search.

   per GitHub releases · cline/cline (L0)

What is not confirmed

The excerpt does not establish cloud-model pricing, comparative quality, or whether changing the cloud selection affects a task already running. It specifies the model choice for new cloud tasks and says that choice is saved separately from local models. [1]

The excerpt does not establish retry timing or billing impact, guarantee access to a GPT-6 deployment on a particular endpoint, or name an alternative to native Google Search for Gemini API keys. Those questions need checks beyond this release note. [1][2][3]

Who should act, who should wait

Act now if you start cloud tasks and need to control their model, or if your team might mistake the missing cloud controls for a fault. Inspect the separate setting and run a new cloud task. Both interface changes shipped in v4.1.26. [1][2]

Test before adopting the update for a workflow that depends on repeated model requests, GPT-6 endpoint routing, or Gemini API-key native Google Search. Wait on a wider rollout if you first need verified retry behavior, deployment availability, or a replacement search path; the supplied material does not establish those details. [1][2][3]

Sources

• v4.1.26 — GitHub releases · cline/cline L0 · primary