--- title: "Cline Cloud Sessions Launch, Default Model Changes: 4-Step Check" description: "Cline v4.1.25 rolls out Cloud sessions, and the default model changed several times from Oct 13 to 17. Here is what changed, plus a 4-step check after you update." url: https://cdnping.org/en/blog/cline-cloud-sessions-default-model-change language: en published: 2026-10-18 category: "Developer tools" tags: ["Cline","Cline Cloud","default model","custom API","AI coding tools"] source: cdnping.org license: Free to quote and summarize with a link to the url above. --- # Cline Cloud Sessions Launch, Default Model Changes: 4-Step Check Cline v4.1.25, released on October 16, adds Cline Cloud sessions. Tasks can run in a hosted sandbox against a GitHub repository and branch you specify. The rollout is gradual, so not everyone sees it right away. In the same week, the default model for the Cline provider changed on October 13, 16, and 17. If you never picked a model by hand, the model you are actually using after an update may not be the one you used last week. Check once after upgrading. Everything below is current as of October 2026 and is based on the release notes on [Cline GitHub Releases](https://github.com/cline/cline/releases). This article does not name specific model versions; go by what your settings show. ## What are Cline Cloud sessions? According to the v4.1.25 release notes, Cline Cloud sessions run tasks in a hosted sandbox, against the GitHub repository and branch you choose. The release notes say it is rolling out gradually, so it is normal if you update and don't see the entry point yet. Per the release notes, a cloud task keeps running after you close your editor. Desktop v0.0.47 (October 16) removed the Cloud sessions toggle from settings, so wherever it has been rolled out it is simply available. The release notes don't say whether it requires payment; check Cline's official docs and your own UI. Other fixes in the same release are worth noting: - Long requests to local models (LM Studio, Ollama) no longer time out. - OpenAI-compatible endpoint URLs are used exactly as you enter them (previously, entering only a domain would silently append `/v1`). - If the provider returns an error midway through a stream, Cline retries automatically up to 3 times. ## How did the default model change this week? Here is the timeline from the release notes (model names omitted): | Date | Version | Change | | --- | --- | --- | | Oct 13 | v4.1.24 | The default model for Cline and OpenRouter was switched to a new default | | Oct 16 | v4.1.25, SDK v0.0.94, CLI v3.0.72, Desktop v0.0.47 | Cline provider default changed again | | Oct 17 | SDK v0.0.95, CLI v3.0.73, Desktop v0.0.48 | Cline provider default restored to the one from Oct 13 | Note the scope: these changes apply to the default for the Cline provider, and to the CLI default when no model is specified. If you explicitly chose a model in settings, or specified one with your own API key, the default change should not affect you, but it is still worth checking. ## How do I confirm which model I'm using after an update? 1. Open Cline settings and look at the API Provider field to see whether it is Cline, OpenRouter, or a custom endpoint. 2. Read the name in the Model field instead of going from memory. If it shows "default" or nothing is explicitly selected, you are following the official default. 3. To pin it, select a model manually. Later changes to the default won't move you. 4. If you use the CLI, specify the model explicitly in the command and don't rely on the no-model default. For teams, add "verify the model after upgrading" to your upgrade steps. A default model change affects answer quality, speed, and your bill. For ways to track price changes, see [How to find out right away when LLM API prices change](/en/blog/track-llm-price-and-model-changes). ## What should I watch for with a custom endpoint URL plus API key? SDK v0.0.95, CLI v3.0.73, and Desktop v0.0.48 (all October 17) include the same fix: on OpenAI-compatible endpoints, requests for newer-generation OpenAI models now send `max_completion_tokens` instead of `max_tokens`. Those requests were previously rejected. Older Azure models still use `max_tokens`. What this means if you use a custom endpoint: - If you connect to the official OpenAI endpoint, Cline handles the request parameters; just update to these versions. - If you connect through a third-party gateway, check whether it accepts `max_completion_tokens`. If you get a 400 error saying the parameter isn't recognized, read the gateway's docs first, then decide whether the gateway needs an adjustment on its side. Support varies, so rely on each service's official documentation. - Since October 16, OpenAI-compatible endpoint URLs are used exactly as typed. Older versions appended `/v1` automatically when you entered only a domain. Per the release notes, a bare-domain URL saved by an older version gets `/v1` appended once so the existing setup keeps working, but for any URL you enter now, make sure the path is correct yourself. For the general approach to configuring endpoint URLs, see [Configuring custom API endpoints for Claude Code, Codex, and OpenCode](/en/blog/configure-custom-api-endpoint-claude-code-codex). When choosing how to connect, compare options in [Official API, cloud provider, or aggregation gateway: how to choose](/en/blog/official-api-vs-cloud-vs-gateway). Cline is a tool that accepts a custom endpoint URL and key. AGI Plan is compatible with both the OpenAI and Anthropic API formats, so you can connect it as described above. AGI Plan is a commercial partner of cdnping; see [agiplan.dev](https://agiplan.dev) for details. It states clearly that it is not an officially authorized channel. ## How does plugin sandboxing affect me? SDK v0.0.94 and CLI v3.0.72 removed the `in_process` plugin mode. Plugins can now run only in a worker thread or a subprocess sandbox. In the CLI, `CLINE_PLUGIN_MODE=in_process` is no longer supported; use `worker` (the default) or `sandbox`. If the host is missing its bootstrap file, plugin loading now fails with an error instead of silently falling back. If you have written or installed a plugin that depends on `in_process`, it may fail to load after the upgrade. Desktop v0.0.47 also fixed an issue in 0.0.46 where marketplace plugins (such as goal) failed to load because the SDK wasn't bundled. Troubleshooting order: update to the latest desktop version first, then check whether the plugin declares `in_process`, and finally read the plugin author's release notes. ## Post-upgrade checklist - [ ] Note the versions before and after the upgrade (desktop, CLI, and SDK separately) - [ ] Verify Provider and Model, and pin them manually if needed - [ ] Custom endpoint: send a small test request and confirm there is no 400 error - [ ] Local models: run one long request and confirm it no longer times out - [ ] Plugins: confirm each one loads correctly - [ ] To watch endpoint stability, see the [API status page](/en/status), or use the [online speed test](/en) to compare first-connection and round-trip times across endpoints Codex also changed its default model in the same week; see [Codex retired its old default model on October 14](/en/blog/codex-model-retirement-cli-0-163) for comparison. ## Sources - [Cline GitHub Releases](https://github.com/cline/cline/releases) (v4.1.24, v4.1.25, SDK v0.0.93 to v0.0.95, CLI v3.0.72/73, Desktop v0.0.46 to v0.0.48)