Home / Gemini 3.6 Flash

Set Up Gemini 3.6 Flash for Claude Code

The exact Claude Code location for a custom base URL and API key is Not yet measured from the supplied source material. The model identifier to pass is gemini-3.6-flash, while the relay endpoint, credential format, and Claude Code configuration path must be verified against the relay documentation or your account panel before use.

Where do you change the base URL and API key in Claude Code?

The exact Claude Code settings file, command, or environment-variable names for changing the base URL and API key are Not yet measured. Do not assume that the URL shown in the relay pricing data is the request endpoint: the supplied source is https://api.openlux.ai/api/pricing, described as the panel's pricing interface, not as a model invocation endpoint.

Before editing Claude Code, obtain these three values from the relay's current account or integration instructions: the chat completion base URL, the API key field expected by the tool, and the model name accepted by the relay. Only the model name is confirmed here: use gemini-3.6-flash. The correct base URL and key-placement method remain Not yet measured.

What should the complete Claude Code configuration look like?

A complete vendor-specific Claude Code file cannot be provided from the supplied facts without inventing a path or field name. Use the following as a fill-in checklist, and replace only values confirmed by the relay and Claude Code documentation: base_url = Not yet measured; api_key = Not yet measured; model = gemini-3.6-flash.

If Claude Code's installed version supports environment variables for these settings, the exact variable names are Not yet measured and should be copied from its documentation rather than guessed. The value for the model field should remain exactly gemini-3.6-flash, including punctuation and capitalization. Do not place a real secret in a shared configuration file or commit it to a repository. The supplied facts do not establish a particular secret-storage mechanism for Claude Code.

How do you verify that the relay configuration is active?

Verification should prove three separate things: Claude Code is reading the intended key, requests are reaching the intended relay endpoint, and the request is selecting gemini-3.6-flash. The exact Claude Code diagnostic command and the relay's request-log location are Not yet measured, so this page cannot responsibly provide a command that may report a false success.

Start with a harmless, identifiable test request and inspect the resulting request record if the relay exposes one. Confirm the selected model string is gemini-3.6-flash rather than a local default or another alias. A successful response alone is insufficient if the client silently falls back to another model. Also confirm that the URL used for inference is not the pricing URL https://api.openlux.ai/api/pricing, which is identified only as the panel's pricing interface.

What Claude Code relay problems should you check first?

The most important configuration risk is mixing up the pricing interface with the model request endpoint. The supplied data confirms that the pricing endpoint exposes panel pricing, but it does not confirm that it accepts chat requests. Treat any response from that URL as pricing data unless the relay's integration documentation explicitly identifies it as an inference endpoint.

A second risk is model-name substitution. The listed model is gemini-3.6-flash, classified as a Google chat model, and no alternate Claude Code alias is supplied. If Claude Code reports an API error after the key is accepted, check the exact model field, the request protocol required by the relay, and whether the account group exposes that model. The supplied facts do not establish Claude Code compatibility, supported request format, error mappings, rate limits, or fallback behavior, so each must be tested rather than assumed.

How should you configure model switching in Claude Code?

Use an explicit model-selection value for each profile, with gemini-3.6-flash as the Gemini profile's exact identifier. The configuration mechanism for multiple Claude Code profiles is Not yet measured, so the profile file name, command-line flag, and environment-variable name must come from the installed Claude Code version or the relay's integration guide.

A practical switching record should contain, at minimum, the profile label, the relay base URL, the credential reference, and the exact model identifier. For the Gemini profile, record model = gemini-3.6-flash. Do not infer that every model listed by the pricing interface can be selected from Claude Code: the interface lists 150 highest-volume models and states that 302 additional models are not shown, but it does not define client-side compatibility or account-level availability.

What is still unknown before using Gemini 3.6 Flash in Claude Code?

The supplied facts confirm the relay's pricing data was pulled at 2026-08-04T16:16:08Z and list gemini-3.6-flash with a base input price of $1.5 per 1 million tokens, a base output price of $7.5 per 1 million tokens, and a cache-hit price of $0.15 per 1 million tokens. These are base prices; the final amount depends on the user's group multiplier. Pricing details belong on the site's pricing page and are included here only to identify the model record, not as a purchase or payment guide.

The following integration facts remain Not yet measured: the Claude Code version covered, the precise base URL, the accepted authentication header or key field, the configuration-file path, the supported request schema, the verification command, error-code behavior, limits, and fallback rules. Record these values from a live integration test before publishing a copy-paste command. This keeps the page useful without presenting an unverified configuration as guaranteed to work.

Still stuck? Full documentation and support are at visit the site.

More on this site

Get started

Check the current pricing record and validate Gemini 3.6 Flash in your integration.

Get a free API key

Official site: OpenLux official site

Last updated 2026-08-05 | Written and maintained by OpenLux.
Latency and pricing figures come from our own measurements. Where they differ from the vendor's site, the vendor's live page wins.