Cline v4.0.11 Adds Claude Opus 5 and Moonshot Kimi K3 Support, Fixes Overstated 1M-Context Pricing

What happened

Cline, the open-source AI coding agent for VS Code and JetBrains, shipped version 4.0.11 on July 24. The release's headline addition is provider support for Claude Opus 5, according to the official changelog, spanning six integration points: Anthropic, Claude Code, AWS Bedrock, Google Vertex, Cline's own provider, and OpenRouter. Support includes what the changelog describes as 1M context window variants of Opus.

The release also adds support for Moonshot Kimi K3, a model from Moonshot AI, within Cline's provider list.

Neither Anthropic nor Moonshot AI has published independent model cards, pricing pages, or benchmark results for Claude Opus 5 or Kimi K3 as of this writing. The only public reference to these models is this Cline changelog entry, so specifics on parameter count, training cutoff, and benchmark performance remain unconfirmed.

Fixes in this release

Two fixes ship alongside the new model support:

• Pricing correction: Cline's cost calculator was overstating charges for Claude Opus 1M-context requests once usage crossed the 200k-token threshold. This is now corrected, meaning any Opus 1M-variant billing estimates displayed in Cline prior to v4.0.11 for large requests should be treated as inaccurate.
• Kimi K3 tool calling: Native tool calling for Kimi K3 models was previously returning empty responses. This is now enabled and fixed, per the changelog.

A minor telemetry change was also included: host plugin version numbers are now transmitted with telemetry events, which helps the Cline team debug issues tied to specific IDE extension builds.

What this means

This is a plumbing release, not a model launch — Cline doesn't train models, it routes requests to them. The substance here is that Cline's engineering team has wired up support for two models (Claude Opus 5 and Kimi K3) that are not yet independently confirmed or documented elsewhere, which is unusual: typically provider integrations follow public model announcements, not precede them. Developers using Cline should treat availability of these models as provisional until Anthropic and Moonshot AI publish their own release details, including pricing, context limits, and benchmark scores.

The pricing bug fix is the more concrete, verifiable item in this release. Teams running high-volume, long-context workloads through Cline's Opus 1M-context path were likely overcharged in cost estimates for any request exceeding 200k tokens — worth checking historical usage reports if cost tracking mattered to your workflow. The Kimi K3 tool-calling fix removes a functional blocker that would have made the model unusable for agentic coding tasks, which is Cline's core use case.

For now, users evaluating whether to switch workloads to Claude Opus 5 or Kimi K3 through Cline should wait for primary-source confirmation from Anthropic and Moonshot AI before making cost or capability assumptions based solely on this changelog.