Cline v4.1.5 リリースCline Releases v4.1.5

Cline, the open-source AI coding agent that runs inside Visual Studio Code, has shipped version 4.1.5, a maintenance-focused update that centers on how the tool communicates about free model availability and how it handles reasoning configuration behind the scenes. The changes are small in scope but address a recurring source of confusion for developers who lean on promotional free models to keep their token costs down.

The headline addition is greater transparency around free model promotions. Cline now explains when a free model promotion is scheduled to end, giving users advance notice rather than leaving them to discover a change only when a request fails. This matters because many of the models exposed through Cline are offered on temporary free tiers, and those arrangements can be withdrawn or converted to paid access with little warning. Surfacing an end date allows developers to plan around the transition, whether that means switching providers, budgeting for paid usage, or finishing a task before access lapses.

Closely related is an improvement to error handling for models that have already been retired. Previously, sending a request to a free model that had been discontinued would produce a generic error accompanied by nothing more than a Retry prompt, an unhelpful loop since retrying a removed model would simply fail again. In v4.1.5, such requests now trigger a dedicated notice that explains the situation and includes a button to pick another model. That shift from a dead-end error to an actionable path is a practical usability gain, reducing the friction of recovering from a model that is no longer available and steering users toward a working alternative instead of leaving them stuck.

On the more technical side, the release changes how reasoning settings are handled across providers. Cline now maps reasoning settings onto a shared path across AI SDK providers, consolidating what was likely provider-specific handling into a common route. Reasoning settings typically control how much intermediate "thinking" a capable model performs before producing an answer, a feature associated with models such as OpenAI's o-series, Anthropic's extended thinking modes, and comparable offerings from other vendors. Unifying this logic across providers should make behavior more consistent regardless of which backend a developer selects, and it tends to simplify maintenance by reducing duplicated code paths. Users are unlikely to see a dramatic change in day-to-day operation, but a shared implementation generally lowers the risk of provider-specific bugs and makes future support for reasoning-capable models easier to add.

For readers unfamiliar with the project, Cline is an autonomous coding assistant that can read and edit files, run terminal commands, and work through multi-step tasks under user supervision. Unlike closed offerings, it is model-agnostic: developers bring their own API keys and can route requests to a wide range of providers, including aggregators that expose many models through a single interface. That flexibility is precisely why free model promotions and retirements have an outsized effect on the Cline experience, and why clearer messaging about their lifecycle is a meaningful quality-of-life improvement.

The update also sits within a broader competitive landscape. Cline operates alongside tools such as GitHub Copilot, Cursor, Aider, and Continue, as well as Roo Code, a community fork that shares much of Cline's lineage. Across this category, vendors and open-source maintainers have been iterating quickly on reasoning support and on the economics of model access, as the underlying model market shifts between free trials, promotional pricing, and paid tiers. Changes like the ones in v4.1.5 reflect that churn: as providers add, promote, and retire models, agent tools have to keep users informed and gracefully redirect them when an option disappears.

Overall, v4.1.5 appears to be an incremental release rather than a feature-defining one, focused on transparency, error recovery, and internal consolidation. The free model promotion notices and the model-picker fallback are the most visible changes for everyday users, while the shared reasoning path is primarily an under-the-hood refinement. Developers relying on free models in particular are likely to notice the improved messaging first, and can consult the project's release notes for the full list of changes accompanying this version.