Cline v4.1.24 defaults to Claude Sonnet 5.5 and adds Claude Haiku 5.5 to its model catalog

Cline v4.1.24 changes the default model for the Cline and OpenRouter providers to Claude Sonnet 5.5 when the user has not chosen one, according to the release notes published on GitHub on 13 Oct. For the Cline provider, Sonnet 5.5 is now the first entry in Cline's recommended list.

The release notes do not include pricing, context window sizes, or benchmark figures for any model mentioned. Those details are not disclosed in the changelog.

Default model changes

• Cline and OpenRouter: Claude Sonnet 5.5 is the fallback when no model is selected.
• Other providers: Defaults change for 302.AI, Cortecs, CrossModel, DevPass, DigitalOcean, Eden AI, GitHub Copilot, Google Vertex AI, LLM Gateway, Meta, Moark, NanoGPT, NEAR AI Cloud, Neuralwatt, Novita AI, Ofox, OpenCode Go, Requesty, Vivgrid, and a few others. Most now point to Claude Haiku 5.5 or Muse Wonder 1.0.
• Impact: Users of those providers who have not pinned a model will get a different default after updating.

Catalog and provider updates

• Adds Claude Haiku 5.5 to the model catalog.
• Adds three providers: AI/ML API, LowRouter, and Prime Inference.
• Adds Step 5 Preview to the Cline free model list.
• The v0 provider now offers only its default model, v0-1.5-md.
• OpenCode Go moves to a new endpoint: https://opencode.ai/inference/go/v1.
• Selecting ai& now shows a dedicated API key and model picker instead of the generic OpenAI-compatible form with base URL and custom headers.

Fixes

Reasoning and thinking

• Turning reasoning off no longer returns a 400 error on GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5, or Claude Opus 5.5. Cline says the fix now works on every provider.
• Level-only Gemini models no longer reject requests with a thinking-budget error.

Context and compaction

• Screenshots and other images no longer inflate the token estimate. Cline says the inflated count caused auto-compaction to trigger well before a conversation approached its limit.
• Auto-compaction now uses the selected model's context window when the catalog entry has none.

Stability and errors

• An MCP server or hook that exits before reading all of its input no longer crashes the extension host about 3 seconds later.
• A 404 from a stale or retired model ID is no longer shown as a sign-in problem. Cline now shows the actual error instead of the logged-out card.
• Errors in the Cline output log keep their details. Production builds previously dropped them, so a line like "Cline OAuth login failed:" appeared without a reason.

Editing and configuration

• Deleting a slash command from the input preserves the spacing of the surrounding text.
• Rule files that are symlinks now load, and skills in .claude/skills are now discovered.
• The editor tool no longer receives null in place of the text to replace from some models.
• Settings migrated from older versions no longer add an empty SAP AI Core provider.

What this means

The default-model change is the item that matters most. Many Cline users never pin a model, so a new default changes their cost and behavior without any action on their part. Cline's own provider and OpenRouter now route unpinned users to Sonnet 5.5, while a long list of aggregators and gateways now lean toward the cheaper-tier Haiku 5.5. Teams that track spend or need reproducible results should pin model IDs explicitly. The changelog gives no per-token pricing, so the cost impact cannot be assessed from this release alone.

The reasoning-off fix matters for a different reason. A 400 error when reasoning is disabled on four separate current models suggests that provider APIs still handle thinking parameters inconsistently. Cline's decision to make the fix work on every provider reflects how much of its work is normalizing those differences.

The compaction fixes address a practical problem. Image-heavy sessions were being compacted prematurely, which degrades agent context for no reason. Falling back to the selected model's context window when the catalog lacks one should also make compaction more predictable on newly added or less-documented models.