Move retention/model training from LLM Gateway overview to retention page & restructure model tables

[devin-ai-integration]

Summary

Addresses feedback that the LLM Gateway overview page is hard to read due to verbose retention/training columns cluttering the model tables. Changes:

1. Removed Retention Policy and Model Training columns from all model tables on the LLM Gateway overview page
2. Restructured model tables from three provider-grouped tables into two unified cross-provider tables:
   • By quality (sorted by LMArena Score, descending)
   • By latency (sorted by latency, ascending; TBD models at bottom)
3. Added a Provider column so models remain identifiable by provider
4. Moved per-provider retention details to the Data Retention and Model Training page under a new "Provider-specific retention policies" subsection
5. Added a Note on the overview page linking to the retention page
6. Moved callouts to after both tables — the deprecation warning, context window note, and retention link now appear after the "By latency" table instead of before the tables (per review feedback from Lee)

Also removed the Description column and the "Unsure which model to choose?" guidance section as part of the table restructuring.

Review & Testing Checklist for Human

• Verify the Description column removal is acceptable. The original tables had a Description column (e.g. "Claude's most intelligent model for building agents and coding") which was dropped in the restructure. The user didn't explicitly request this — confirm this info loss is OK or should be restored.
• Verify the "Unsure which model to choose?" section removal is acceptable. This guidance ("Consider Claude models for nuanced reasoning…") was also removed. Confirm it's not needed or should be kept.
• Verify sort order and model completeness in both tables — all 21 models should appear in each table, LMArena scores descending in the first, latency ascending (TBD last) in the second.
• Check rendered output via Fern preview deployment to confirm tables render cleanly, callouts appear after both tables, and the anchor link (#llm-gateway-production-environment) on the overview page resolves correctly to the retention page section.
• Review the moved retention content on the data retention page to ensure accuracy is preserved (especially the Bedrock failover note for Anthropic and the OpenAI open-weight vs API model distinction).

Notes

• Requested by @dan-ince-aai
• Callout reordering requested by Lee during PR review
• Link to Devin Session

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[github-actions]

🌿 Preview your docs: https://assemblyai-preview-9c371b3c-8bbe-4f7f-b49e-2a82a137c6da.docs.buildwithfern.com/docs

[github-actions]

🌿 Preview your docs: https://assemblyai-preview-ae36392f-2eec-4ff3-9fc6-428ccc41eea7.docs.buildwithfern.com/docs

[github-actions]

🌿 Preview your docs: https://assemblyai-preview-921009e7-3185-4e95-82c0-c41284c28fc1.docs.buildwithfern.com/docs