[AI] Refresh model catalogs

[superhighfives]

Summary

Refreshes the Workers AI and unified catalog model data from the production APIs.

Documentation checklist

• The change adheres to the documentation style guide.

[github-actions]

This pull request requires reviews from CODEOWNERS as it changes files that match the following patterns:

Pattern	Owners
/src/content/catalog-models/	@abhishekkankani, @ethulia, @kathayl, @kflansburg, @mattrothenberg, @mchenco, @palashgo, @roerohan, @shridhar-cf, @superhighfives, @thebongy, @zeke, @cloudflare/product-owners, @cloudflare/content-engineering
/src/content/workers-ai-models/	@craigsdennis, @ethulia, @kflansburg, @mchenco, @shridhar-cf, @superhighfives, @cloudflare/product-owners

[cloudflare-workers-and-pages]

🚀 Deploying Preview to Cloudflare 🚀

Preview URL: https://superhighfives-refresh-ai-model-catalogs.previews.developers.cloudflare.com (commit 70ed4d6)

This URL reflects your latest Preview deployment

Preview Deployments by commit

Status	Deployment URL	Commit	Updated (UTC)	See this deployment's details
Build: Success ✅ Deployment: Success ✅ View logs ↗	https://ea46129f.previews.developers.cloudflare.com	70ed4d6	2026-09-14T16:47:49.145Z	Visit the dashboard ↗

[github-actions]

Preview URL: https://70ed4d6b.preview.developers.cloudflare.com
Preview Branch URL: https://superhighfives-refresh-ai-model-catalogs.preview.developers.cloudflare.com

[cloudflare-docs-bot]

Review

Review in progress for commit 70ed4d6…

[superhighfives]

/bonk review

[ask-bonk]

These pricing keys render verbatim in the comparison table (formatModelPricing in src/util/models/model-format.ts), unlike the Input … (per 1M) keys above them — formatPricingLabel only normalizes the … tokens (per 1M) pattern, so output_text_tokens/output_video_tokens show as raw labels ("output text tokens: $9.00"). The same issue applies to the other new omni file and google-gemini-3.1-flash-tts.json.

Suggested change

	"output_text_tokens": 9,
	"Reasoning tokens (per 1M)": 9,
	"output_video_tokens": 17.5,
	"Default (per second)": 1.5
	"Output text tokens (per 1M)": 9,
	"Reasoning tokens (per 1M)": 9,
	"Output video tokens (per 1M)": 17.5,
	"Default (per second)": 1.5

[ask-bonk]

Posted 1 inline suggestion.

LOW — google-gemini-omni-1.1-flash.json (and the identical pricing in google-gemini-omni-flash.json and google-gemini-3.1-flash-tts.json): the new pricing objects mix nicely-formatted labels (Input text tokens (per 1M)) with raw snake_case keys (output_text_tokens, output_video_tokens, output_audio_tokens) and Default (per second). formatModelPricing in src/util/models/model-format.ts only normalizes the … tokens (per 1M) pattern, so the snake_case keys render verbatim in the comparison table (e.g. "output text tokens: $9.00"). Suggest normalizing the keys (or extending formatPricingLabel).

Out of scope: this refresh removes 22 legacy Workers AI models (e.g. llama-3-8b-instruct, llama-3.1-8b-instruct, mistral-7b-instruct-v0.1) — no in-repo references to those pages exist, so the build is unaffected, but external URLs to /workers-ai/models/<model>/ will now 404; consider redirects if users should migrate to the catalog equivalents.

Everything else validated: all 241 changed JSON files parse, required schema fields present, model_id/legacy id unique, new files follow naming conventions, providers resolve in authorData, and no dangling references anywhere in src/ or public/.

github run