[Enhancement]: Add Claude Opus 4.6 model support

[saneroen]

What features would you like to see added?

Add Claude Opus 4.6 model to the available model configuration.

Claude Opus 4.6 is the latest version in the Opus 4.x series and should be available for users to select in LibreChat. The model supports 64K max output tokens and uses the same pricing structure as Opus 4.5.

Models to add:

• claude-opus-4-6 to sharedAnthropicModels array

Files to update:

• packages/data-provider/src/config.ts - Add model to array
• api/models/tx.js - Add pricing configuration (matching opus-4-5)

More details

Background

Claude Opus 4.6 is the latest version in Anthropic's Opus 4.x model series, offering improved performance and capabilities for complex reasoning tasks. It should be available alongside other Opus 4.x models (opus-4-1, opus-4-5) in the model selection.

Implementation Details

Model Configuration:

• Model name: claude-opus-4-6
• Max output tokens: 64K (handled automatically via existing regex patterns in schemas.ts)
• Pricing: Same as opus-4-5 (prompt: 5, completion: 25, cache: write 6.25, read 0.5)

Files to Modify:

1. packages/data-provider/src/config.ts

   • Add claude-opus-4-6 to sharedAnthropicModels array (after claude-opus-4-5)

2. api/models/tx.js

   • Add pricing entry in tokenValues object
   • Add cache pricing entry in cacheValues object

Expected Behavior

After implementation, claude-opus-4-6 should:

• Appear in the model selection dropdown for Anthropic endpoint
• Support 64K max output tokens (via existing regex patterns)
• Use correct pricing for token calculations
• Work seamlessly with existing model infrastructure

Testing

• Verify model appears in UI
• Confirm 64K token limit works (existing regex should handle this)
• Test pricing calculations
• Run build and lint checks

References

• Anthropic API documentation
• Existing Opus 4.x model implementations in codebase

Which components are impacted by your request?

General

Pictures

Direct Changes

• packages/data-provider/src/config.ts - Add model to sharedAnthropicModels array
• api/models/tx.js - Add pricing configuration

Automatically Updated (No Code Changes)

• Backend API - Model fetching services will automatically include the new model
• Frontend UI - Model selection dropdowns will automatically show the new model
• Settings panels - Model configuration options will automatically include it

No Changes Needed

• Token limits - Already handled via regex patterns (opus-4-6 matches 4.5+ pattern for 64K)
• Test files - Already have tests for opus-4-6 in schemas.spec.ts
• Bedrock models - Not needed (opus-4-5 also not in Bedrock list)
• Vertex AI - Uses different naming conventions

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

Add the model via .env, no code changes needed:

ANTHROPIC_MODELS=claude-opus-4-6,claude-haiku-4-5

main thing that will need to be added still is 1M context window support for this model.

[saneroen]

Thanks for the feedback. I will update the configuration to add 1M context window support for claude-opus-4-6:
Context window: 200000 tokens and 1M tokens in beta
Max output tokens: 128,000 tokens

[dirkpetersen]

Or for those who use AWS Bedrock ....

BEDROCK_AWS_MODELS=global.anthropic.claude-opus-4-6-v1,......

I noticed that if I just use this as is .... with the latest LibreChat as of this morning .... it can reason itself to almost to death before it comes up with a result .... which is then quite nice actually ... but I am not sure if the default setting should be that we have to wait for 2-5 minutes watching many many lines of reasoning. I was using artifacts