[Enhancement]: Add Gemini 3.6 Flash and Gemini 3.5 Flash-Lite support

[marlonka]

What features would you like to see added?

Please add first-class support for Google's newly generally available models:

• Gemini 3.6 Flash — gemini-3.6-flash
• Gemini 3.5 Flash-Lite — gemini-3.5-flash-lite

The models should be available through both Google integrations supported by LibreChat:

1. Gemini API, authenticated with a Google AI Studio API key
2. Google Cloud / Gemini Enterprise Agent Platform (formerly Vertex AI)

Both models were released on July 21, 2026 and are described by Google as production-ready. Gemini 3.6 Flash targets agentic, coding, multimodal, and computer-use workloads, while Gemini 3.5 Flash-Lite targets high-throughput, low-latency execution such as extraction, classification, structured data processing, and subagent work.

More details

Model metadata

Model	Model ID	Default thinking level	API pricing
Gemini 3.6 Flash	gemini-3.6-flash	medium	$1.50 / 1M input, $7.50 / 1M output
Gemini 3.5 Flash-Lite	gemini-3.5-flash-lite	minimal	$0.30 / 1M input, $2.50 / 1M output

Google documents a 1M-token context window, up to 64k output tokens, thinking support, multimodal support, function/tool use, and native Computer Use support for both models.

Important API compatibility changes

At minimum, the LibreChat integration and documentation should explicitly account for the API changes Google introduced with these models and link to Google's documentation as the source of truth:

• temperature, top_p, and top_k are deprecated and ignored for these models; future model generations will return HTTP 400 when they are supplied.
• Prefilled model turns are no longer supported. Requests ending in a non-empty model role turn return HTTP 400.
• thinking_level should be used instead of thinking_budget where applicable.
• candidate_count is unsupported for Gemini 3.x.
• Existing function-calling and thought-signature requirements for Gemini 3.x still apply.

LibreChat should therefore avoid forwarding deprecated or unsupported generation parameters for these model IDs and ensure conversation payloads do not rely on prefilled assistant/model turns. This is especially important because a model can appear selectable while still failing at runtime due to provider request serialization.

Official API changes and migration guidance:
https://ai.google.dev/gemini-api/docs/latest-model#api-changes-and-parameter-updates

Suggested scope

• Add both model IDs to the Gemini API / Google AI Studio model configuration and model-selection UI.
• Add both model IDs to the Google Cloud endpoint integration used for Gemini models.
• Apply model-specific parameter filtering or request normalization for the new API rules.
• Preserve configurable thinking levels:
  • Gemini 3.6 Flash: medium by default, optionally high
  • Gemini 3.5 Flash-Lite: minimal by default, optionally medium or high
• Verify standard chat, multimodal input, structured output, function calling/tool use, streaming, and multi-turn conversations.
• Document any currently unsupported native capabilities, such as Computer Use, separately rather than blocking basic model availability.

Acceptance criteria

• gemini-3.6-flash and gemini-3.5-flash-lite can be configured and selected with a Gemini API / Google AI Studio API key.
• Both models can be configured and selected through LibreChat's Google Cloud Gemini endpoint.
• Basic chat and streaming requests complete without deprecated sampling parameters being sent.
• Multi-turn conversations do not produce prefilled-model-turn HTTP 400 errors.
• Thinking-level configuration is translated correctly for each model.
• Tool/function calling and multimodal requests are covered by integration tests where the existing endpoint supports them.
• LibreChat documentation links to Google's current migration/API-changes page as the source of truth.

Official sources

• Announcement and availability: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/
• Model overview, IDs, pricing, capabilities, and migration guidance: https://ai.google.dev/gemini-api/docs/latest-model
• Gemini API model documentation: https://ai.google.dev/gemini-api/docs/models

Which components are impacted by your request?

• Endpoints / provider integrations
• Model configuration and discovery
• Model selector UI
• Request parameter normalization
• Thinking configuration
• Function calling / tools
• Multimodal support
• Documentation and tests