Price value
Lower blended input/output price per one million USD tokens scores higher within the current snapshot.
AI model value leaderboard
Choose the right model for the work.
A transparent ranking of currently available routes by price, affirmative capability evidence, and known limits. It helps you compare models quickly without pretending one score can measure every kind of intelligence.
Catalogue-backed ranking
Scores are recomputed from the approved public snapshot. Unknown evidence stays unknown.
Live catalogue unavailable
The value board is waiting for a current snapshot.
InferencePass does not show stale prices or availability as current. Try again shortly, or browse the model catalogue.
Methodology
A value score you can inspect.
The ranking is designed to make price and evidence visible—not to turn a catalogue into a fake benchmark.
Capability coverage
Affirmatively supported text, vision, tools, structured output, reasoning, streaming, and caching capabilities count. Unknown is zero.
Known limits
Known context-window and maximum-output values contribute. Missing limits do not receive an estimate.
What this board does not measure: It does not measure model intelligence, benchmark quality, or first-party identity. Test the routes on your own prompts before using them in production.
Questions
How the leaderboard works.
What does the InferencePass AI model leaderboard measure?
The leaderboard measures catalogue-backed value: 45% normalized input/output price, 35% affirmative capability coverage, and 20% known context and output limits. It is not an intelligence benchmark and does not claim that the top-ranked route is universally the best model.
How is the value score calculated?
Eligible models are ranked using exact catalogue prices per one million USD tokens, affirmative capability evidence, and known context/output limits. Lower blended input/output price scores higher. Unknown capability evidence contributes zero rather than being treated as supported.
How often does the leaderboard update?
The page reads the current approved catalogue snapshot and shows its version and generated timestamp. A new snapshot can change prices, availability, capabilities, and rank order.
Are these first-party OpenAI, Anthropic, Google, or GLM APIs?
No. InferencePass is an independent service. Provider and model names are informational, and a compatibility-labelled route does not prove first-party identity, endorsement, or permission.
One endpoint, many options
Compare the shortlist on your prompts.
Use the Playground to test outputs, or create an account to keep your model choice flexible behind one compatible API.