Cline API request fails after a model switch? Verify the selected model.

A model change can affect provider routing and request compatibility. Preserve the task diff, then check what model Cline actually selected before retrying the same request.

Do this first: Record the Cline CLI version, provider, configured model, and model shown in the failed request. Compare those values with issue #6936 only if you are using a matching CLI setup.

What the failure looks like.

After switching models in Cline CLI, a request returns API Request Failed and the selected model may not match the configured model. Issue #6936 describes this on Cline CLI 3.33.1 with the OpenAI Native provider.

Why it happens.

The report is CLI-specific and closed without a documented workaround or shipped fix in its body. It does not prove the same issue affects the VS Code extension or current versions.

Fix it without losing the task.

Record the Cline CLI version, provider, configured model, and model displayed for the failed request.

Save the exact redacted error and request timestamp. Do not include API keys or private source code in an issue report.

Compare against one known-good provider/model configuration in a fresh session, changing only one variable. Keep the project task unchanged.

If the selected model differs from configuration, attach the evidence to the upstream report. Verify the worktree before replaying the request.

What the next AI should receive.

Cline CLI version, provider, configured and reported model names, exact redacted error, whether the request reached the provider, current diff, and one known-good comparison.

First safe action

Do not repeatedly switch models in the same session. Capture the selected-model display and redacted request error, then compare one known configuration in a fresh session.

Evidence boundary

Issue #6936 is a single historical report for Cline CLI 3.33.1. It does not establish a current regression, explain every API Request Failed response, or cover the VS Code extension.

Diagnostic checkpoint

The useful evidence is the model actually sent, not only the value shown in settings. Keep the CLI and IDE extension cases separate.

How ShardStitch helps.

ShardStitch can carry the model/provider decision, failed request, diff, and next verification into a clean session. It cannot correct provider routing or infer the model sent over the network.

Source trail.

Specific report, checked September 24, 2026: Cline CLI issue #6936 was opened October 17, 2025 for Cline CLI 3.33.1. It reports the OpenAI Native provider using a different model than the configured one and then returning “API Request Failed.” The issue is closed, but its body does not document a user workaround or the shipped fix. Scope matters: this is a CLI report, not proof that the same defect affects the VS Code extension. Browse the project issue tracker.

FAQ.

Does Cline issue #6936 apply to the VS Code extension?

The report is for Cline CLI 3.33.1 and does not establish the same behavior in the VS Code extension. Treat the surfaces separately.

Did issue #6936 document a workaround or confirmed fix?

Its body does not document a user workaround or the shipped fix. Verify current CLI/provider behavior rather than assuming a historical report describes the current version.

Scope note: This guide addresses one reported Cline CLI model-selection/request failure. The issue does not document a verified fix; recheck current versions and provider behavior.