Cline with GPT-6: the release switches max_tokens to max_completion_tokens
The official release states that GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints; before that, such requests were rejected, and the reason is in the parameter name the client sends. As of 2026-10-20, the official release names CLI 3.0.73 and Desktop 0.0.48 as the versions that say so.
Updated 2026-10-20
One key for Claude, GPT and leading Chinese models, billed per token, up and running in minutes.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
What matters: parameter name, the versions the notes name, endpoint rules, QCode setup
the parameter name Cline sends changed
The parameter name the official release names is max_completion_tokens — GPT-6 models expect that name on OpenAI and Azure OpenAI-compatible endpoints.
the versions the official notes list as fixed
Cline CLI 3.0.73 and Desktop 0.0.48 were both published on 2026-10-17, and the first line of both release notes reads the same: GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints.
the endpoint rules
The official sentence is limited to OpenAI and Azure OpenAI-compatible endpoints. The QCode docs say: GPT models go through the OpenAI-compatible endpoint, and Claude only through the Anthropic endpoint, with the address in the QCode section below.
two ways to connect
Following the QCode docs page Cline integration: way A picks Provider Anthropic and ticks Use custom base URL (for Claude models), way B picks Provider OpenAI Compatible (for GPT models); the field values are in the QCode section below.
how the official release describes the error
The official Cline release gives no error text, only the reason: earlier Cline builds sent max_tokens instead of max_completion_tokens, so every request was rejected. The notes carry no HTTP status code, error code or error body, and name no earlier affected version range.
the official release updated two versions on 2026-10-17
Checked on 2026-10-20: Cline CLI 3.0.73 and Desktop 0.0.48 were both published on 2026-10-17, and both lead with the line that GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints. Each release also adds a line about the default model: the CLI release says the Cline provider uses Claude Sonnet 5.5 again when no model is chosen, the desktop release says the Cline provider default model is Claude Sonnet 5.5 again.
timeline
The QCode docs page Cline integration cites the Cline CLI 3.0.69 release notes (2026-10-07).
The QCode docs page Cline integration was updated on 2026-10-15 and states that Cline offers two ways to connect: Anthropic and OpenAI Compatible, each with its own base URL.
Cline CLI 3.0.73 and Desktop 0.0.48 shipped on the same day, and the official notes in both releases say that GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints.
what the official notes say and what they leave out
confirmed (word for word from the official release)
① The official wording, the same sentence word for word in CLI 3.0.73 and Desktop 0.0.48: every request was rejected because Cline sent max_tokens instead of max_completion_tokens. ② The first half of the same sentence says GPT-6 models work on OpenAI and Azure OpenAI-compatible endpoints. ③ Both releases were published on 2026-10-17. ④ The official notes give no status code, error code or error body for the rejected requests.
what the official notes do not publish
The official release does not say: the HTTP status code, error code and error body of the rejected requests; which earlier Cline versions were affected.
two options: upgrade Cline, or switch model first
upgrade Cline
The official release states that in CLI 3.0.73 and Desktop 0.0.48 GPT-6 models can be used on OpenAI and Azure OpenAI-compatible endpoints. If you already work with GPT-6, check your own Cline version first.
use another model for now
The Cline provider uses Claude Sonnet 5.5 again when no model is chosen, and the official release puts it back at the top of the recommended list. In the QCode docs the daily default is claude-sonnet-5, claude-opus-5 is for hard tasks, and claude-sonnet-4-6 and claude-opus-4-8 are still sold. With those models you do not have to deal with the GPT-6 endpoint requirement for now.
how to check and how to configure
① Check the Cline version first: CLI 3.0.73 and Desktop 0.0.48 are the versions the official release lists as carrying this fix, published 2026-10-17. ② Then check the endpoint: GPT models go through the OpenAI-compatible endpoint, with base URL and Model ID filled in as the QCode docs page Cline integration describes, with the values in the QCode section below. ③ Do not put a Claude model into the OpenAI-compatible endpoint; the QCode docs say it returns model_not_available_on_endpoint. ④ Send one short message in the Cline chat box to verify the setup; the interface shows the token usage and the estimated cost of every request.
connecting Cline to QCode for GPT-6
Following the QCode docs page Cline integration (updated 2026-10-15): one API key (prefix cr_) serves both the Anthropic and the OpenAI protocol. Way A picks Provider Anthropic, ticks Use custom base URL and fills in https://api.qcode.cc/api for Claude models (claude-sonnet-5 or claude-opus-5); way B picks Provider OpenAI Compatible, sets base URL https://api.qcode.cc/openai/v1, uses the key that starts with cr_ and puts gpt-6-sol into Model ID for GPT models. The OpenAI-compatible endpoint does not serve Claude models, and a claude-* model id returns model_not_available_on_endpoint. gpt-6.1-sol is also available on QCode, the model list is on qcode.cc/models; billed per token, per-model rates are listed on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 to 90 days.
frequently asked questions
What error does Cline show with GPT-6?
The official release says every request was rejected because the parameter name Cline sent was max_tokens instead of max_completion_tokens.
Which Cline versions are documented as supporting GPT-6?
Cline CLI 3.0.73 and Desktop 0.0.48, both published by the official release on 2026-10-17. The official notes do not list an earlier affected version range.
How do the Cline version and the endpoint correspond?
The official release names CLI 3.0.73 and Desktop 0.0.48 as the versions carrying this fix. On the QCode side GPT models go through the OpenAI-compatible endpoint, and Claude only through Anthropic. The full OpenAI Compatible setup is on the site's /cline-openai-compatible-guide page.
Can I call Claude through the OpenAI-compatible endpoint?
No. The QCode docs state that the OpenAI-compatible endpoint serves only the GPT family and the four Chinese model families, and that a claude-* id returns model_not_available_on_endpoint; for Claude use the Anthropic option with base URL https://api.qcode.cc/api.
Which GPT-6 models can I call on QCode?
gpt-6-sol and gpt-6.1-sol are both available on QCode, and the full list sits on qcode.cc/models; QCode bills per token and the per-model rates are on /models.
How does QCode bill?
QCode charges per token and the per-model rates sit on /models. The same key that starts with cr_ works for both the Anthropic and the OpenAI protocol, and the tools share one key.
sources
Official wording: the Cline release notes on github.com/cline/cline/releases (CLI 3.0.73 and Desktop 0.0.48, published 2026-10-17, fetched and checked on 2026-10-20). On the QCode side: the Cline integration page (updated 2026-10-15) and the endpoints and API formats page (updated 2026-10-19) on docs.qcode.cc, plus the pay-as-you-go page on qcode.cc in five languages (fetched and checked on 2026-10-20).
connect Cline to QCode
Billed per token, per-model rates are listed on /models; gpt-6-sol and gpt-6.1-sol are available on QCode.
related reading
Claude Code Pay-as-you-go Packs: No Subscription, 7–90 Days
One-time quota packs valid for 7–90 days, no subscription; extra packs add to the same key
Cline OpenAI-Compatible API Setup 2026: BYOK & Model IDs
the three fields that connect Cline to an OpenAI-compatible endpoint, and how to choose between your own key and a $20 subscription
GPT-6.1 Sol API Guide: Pricing, 272K Rule and vs gpt-6-sol
The 272K context billing rule and a line-by-line price comparison with the other model of the same family.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
Verified on 2026-10-20 against the official Cline releases and the QCode docs; version behavior changes over time, so follow the official release. Model availability on the QCode side is listed on /models.