feat: update EmpirioLabs models in model_prices_and_context_window.json

[Adam-Dalloul]

Refreshes EmpirioLabs price and context-window metadata from the public /v1/models catalog. Scope: chat, embedding, rerank models, keyed empiriolabs/.

[greptile-apps]

Greptile Summary

Refreshes the EmpirioLabs model registry from its model catalog

• Adds chat, embedding, and rerank model identifiers
• Records pricing, context and output limits, cache pricing, and supported capabilities
• Keeps each model keyed under the empiriolabs/ provider namespace

Confidence Score: 5/5

The PR appears safe to merge because no concrete failure was identified in the added model metadata

The new records remain valid top-level model entries, use supported registry modes, and no added price, limit, or capability value was shown to violate a reachable LiteLLM contract

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds structurally valid EmpirioLabs model metadata; no concrete changed-code defect was established

_{Reviews (1): Last reviewed commit: 7a844f4 | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[devin-ai-integration]

Reviewed while consolidating registry changes into the rolling PR #37902. Not rolling this one in yet, reason in that PR description under 'Source PR disposition for this run'. Leaving it open for the author to respond.

[Adam-Dalloul]

Thanks for the flag in #37902. I removed the entries that were priced at zero and regenerated the rest.

The zeros were products a per-token schema cannot express: research and search answers billed per call, the agent runner billed per task, image generation billed per image, the text detector billed per 1,000 words, and a few chat models on a fixed per-message price. LiteLLM meters per-token and per-second rates and has no input_cost_per_request path, so leaving those in would have made it report paid requests as free. Better absent than wrong.

The 92 entries left all have a real per-token rate, and every price, context window and cache rate in the diff was generated from https://api.empiriolabs.ai/v1/models, which is public and needs no API key, so it should be straightforward to verify. supports_prompt_caching is set only where the cached input rate is actually below the prompt rate.