Last refreshed 2026-05-11. Next refresh: weekly.
Pick the integration path by priority first, then drill into model, host, price, benchmark, and freshness evidence before you commit code.
Decision strip
Live shortlist
Frontier quality
Highest GPQA route in tracked API data
Use this when benchmark headroom matters more than token price.
$/1M input
$2.00
$12.00 output
Google-Proof Q&A
94.3%
tracked signal
Price/performance
Best benchmark-per-input-dollar route
Use this when you need credible quality without letting input cost run away.
$/1M input
$0
$0 output
Google-Proof Q&A / input $
58,600x
tracked signal
OAI-compatible alternative
Top non-OpenAI host with chat-completions evidence
Use this when a base URL swap is attractive, then verify IDs and headers.
$/1M input
$0.950
$4.00 output
Google-Proof Q&A
90.5%
tracked signal
Browse by intent
Start with the capability-ranked API leaderboard when quality is the first constraint.
Compare first-party, hyperscaler, aggregator, and fast-inference hosts by tracked model count.
Use paid API price rows when cost pressure matters more than absolute benchmark rank.
Choose models with stronger tool-use, multi-turn, and coding-agent signals.
Find function-calling and structured-output candidates for production control loops.
Route RAG, document, and codebase workloads toward larger tracked context windows.
These checks prevent model-level capability from being mistaken for provider-route capability.
Confirm the provider route supports streaming responses before you design UI latency around it.
Prefer models with tool-use or structured-output evidence when the app calls your own functions.
Use batch-price evidence for offline enrichment, eval generation, and cost-sensitive backfills.
Check model IDs, base URLs, auth headers, and SDK support before treating a provider swap as drop-in.
| Capability | What to verify | Concept |
|---|---|---|
| Streaming | Confirm the provider route supports streaming responses before you design UI latency around it. | Route docs |
| Function calling | Prefer models with tool-use or structured-output evidence when the app calls your own functions. | Open |
| Batch APIs | Use batch-price evidence for offline enrichment, eval generation, and cost-sensitive backfills. | Open |
| OAI-compatible routing | Check model IDs, base URLs, auth headers, and SDK support before treating a provider swap as drop-in. | Route docs |
We maintain pricing, benchmarks, and freshness for every model on this site. Want it as a JSON feed in your tooling?