Model Not Found: Why the Same Model Has Different IDs on Each Aggregator

Why Model IDs Differ Across Aggregators

When you switch between LLM aggregators, you may notice that the same underlying model has a different string identifier on each platform. A request that works on one service can return a 404 or silently fall back to a different model on another. This is not a bug—it stems from how aggregators manage model naming.

Each aggregator typically defines its own internal model IDs to handle provider-specific variations, version updates, and routing logic. For example, one platform might use claude-3-5-sonnet-20240620, while another uses claude-3-5-sonnet-latest or a custom alias like claude-sonnet. These differences exist because:

• Provider aliases vary: Model providers often expose multiple identifiers for the same model (e.g., a dated version and a latest tag). Aggregators choose which one to surface.
• Versioning schemes differ: Some platforms pin to specific snapshot dates, while others default to the newest stable version.
• Routing and fallbacks: Aggregators may map a generic name to different backends based on availability, cost, or region.
• Custom naming: To avoid confusion or trademark issues, some aggregators create their own model names.

Common Symptoms of a Wrong Model ID

If you use a model ID from another platform, you might encounter:

• 404 Model Not Found: The aggregator does not recognize the string.
• Silent fallback: The request succeeds but returns output from a different model, often with degraded quality or unexpected behavior.
• Error messages about invalid model: Some APIs return a 400 error with a list of supported models.
• Unexpected billing: If fallback occurs, you might be charged for a model you did not intend to use.

These issues are frustrating, but they are easy to avoid once you know how to find the correct model ID for your aggregator.

How to Find the Canonical Model ID on This Site

This platform uses its own model IDs to ensure consistent routing and billing. To find the correct ID:

1. Check the models page: Visit the models section in your dashboard or the public models list. This is the single source of truth for available model IDs.
2. Use the API endpoint: If you have an API key, you can query the /models endpoint (or equivalent) to get a programmatic list of supported models.
3. Search the documentation: Our docs include a model reference table with canonical IDs and their corresponding providers.
4. Look for aliases: Some models have multiple acceptable IDs (e.g., a short alias and a full version). The models page will indicate if aliases are supported.

Always copy the model ID directly from our list rather than reusing an ID from another platform.

Best Practices to Avoid Model ID Errors

• Do not hardcode model IDs from other aggregators: Treat each platform as having its own namespace.
• Use environment variables or config files: Store model IDs in a central place so you can update them easily.
• Validate model IDs in development: Before deploying, test your requests against the target aggregator.
• Check for version updates: Model IDs can change when providers release new versions. Subscribe to changelogs or periodically re-check the models page.
• Handle errors gracefully: Implement fallback logic that catches 404s and logs the invalid model ID for debugging.

What to Do Before Filing a Ticket

If you get a "model not found" error, follow these steps before contacting support:

1. Verify the model ID against our official models list.
2. Check for typos: Model IDs are case-sensitive and may include hyphens or underscores.
3. Try the latest alias: If you used a dated version, see if a latest alias exists.
4. Test with a minimal request: Use curl or a simple script to isolate the issue.
5. Review recent changes: Model IDs may have been deprecated or renamed.

If the issue persists after these checks, include the exact model ID, the full error message, and the timestamp in your ticket. This helps our team resolve the issue quickly.

Why This Matters for Your Workflow

Using the correct model ID ensures you get the model you expect, with predictable performance and billing. It also prevents silent fallbacks that can skew evaluation results or degrade user experience. By treating model IDs as platform-specific and verifying them against our canonical list, you avoid unnecessary debugging and support cycles.

Remember: when in doubt, check our models page first. It is the authoritative source for model IDs on this platform.