Find models and compare providers

Find models that gateways list in their catalogs. Availability can depend on credentials, region, deployment, and account access. We cite provider claims and do not test inference.

Compare model families, versions, aliases and provider routes before interpreting catalogue totals. Use an approved-model inventory for migration contract tests.

Provider source coverage

Public catalogs and compatibility lists describe different things. A source listed here may still be awaiting collection and review. Unknown coverage does not mean unsupported.

• Eden AI — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• Hugging Face Inference Providers — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• OpenRouter — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• Requesty — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• Braintrust Gateway — Public catalog. Official source

  Strict literal-field parser of public supported-model registry. Costs are upstream references; account keys and gateway billing determine actual charges.

• Helicone — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• Kong AI Gateway — Integration documentation only. Official source

  We found integration documentation, but no exhaustive public model inventory. Missing listings do not mean unsupported.

• Merge Gateway — Public catalog. Official source

  Imported public source subset; prices may be upstream references or unavailable. Account access and configuration requirements still apply.

• Orq.ai Router — Public catalog. Official source

  Public supported-model tables across API surfaces, deduplicated by exact ID. Complete public pricing feed not established; prices remain unknown.

• Portkey — Compatibility or model defaults. Official source

  Public per-provider cost registry. Defaults excluded; cents/token converted to USD/million tokens. Upstream reference prices, not all-inclusive Portkey charges.

• Vercel AI Gateway — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• agentgateway — Compatibility or model defaults. Official source

  Exact IDs from official provider cookbook; documented compatibility requiring configuration. No universal gateway token price.

• Apache APISIX AI Gateway — Integration documentation only. Official source

  We found integration documentation, but no exhaustive public model inventory. Missing listings do not mean unsupported.

• Bifrost — Compatibility or model defaults. Official source

  Public datasheet, filtered to chat/completion/embedding modes. Upstream cost references; registry membership alone does not establish configured availability.

• Envoy AI Gateway — Integration documentation only. Official source

  We found integration documentation, but no exhaustive public model inventory. Missing listings do not mean unsupported.

• Higress — Integration documentation only. Official source

  We found integration documentation, but no exhaustive public model inventory. Missing listings do not mean unsupported.

• LiteLLM — Compatibility or model defaults. Official source

  Imported entries are chat models in LiteLLM’s published registry. They describe compatibility metadata, not hosted availability; configured providers and credentials determine access.

• LLM Gateway — Public catalog. Official source

  A public model catalog is available. Account access, region, credentials, and deployment requirements may limit availability.

• MLflow AI Gateway — Integration documentation only. Official source

  We found integration documentation, but no exhaustive public model inventory. Missing listings do not mean unsupported.

• New API — Compatibility or model defaults. Official source

  Official channel ModelList defaults, labeled partial compatibility. Instance-specific pricing ratios are not imported as universal prices.

• Amazon Bedrock — Public catalog. Official source

  Official model-card subset; token prices only where an unambiguous match exists in public us-east-1 pricing. Other regions and non-token prices are not imported.

• Azure AI Foundry — Public catalog. Official source

  Documented Azure-direct model IDs; retail token pricing limited to exact matches in eastus2/global. Deployment required. Other regions and compute prices are not imported.

• Cloudflare AI Gateway — Public catalog. Official source

  Public model catalog across Workers AI and third-party routes. Only explicitly published USD/million token input/output prices are imported; other units remain unknown.

• Google Vertex AI — Public catalog. Official source

  Documented managed-model subset from Google model pages. Token prices only for exact matches in supported pricing tables; self-deployed and non-token charges excluded.

• TrueFoundry AI Gateway — Compatibility or model defaults. Official source

  Imported public source subset; prices may be upstream references or unavailable. Account access and configuration requirements still apply.

• Fireworks AI — Public catalog. Official source

  Published serverless pricing subset. Named Fast, region, Standard and Priority tiers preserved; not an exhaustive dedicated-deployment model inventory.

• Groq — Public catalog. Official source

  The imported list covers production text models. Preview models, audio models, and compound systems are outside this initial collection.

• Together AI — Public catalog. Official source

  The imported list covers the published serverless chat table. Dedicated deployments and other modalities have separate catalogs.

• Respan — Public catalog. Official source

  Active chat models from Respan’s public catalog. Some listings require your own provider credentials. Prices are published list rates, before account discounts and other fees.

• AI Gateway HQ — Integration documentation only. Official source

  Provider integrations are documented, but no exhaustive public model inventory is available. Configured aliases and managed Bedrock eligibility depend on the workspace.

• Velokey — Integration documentation only. Official source

  Documented model integrations; exact available IDs require an account. No model-price inventory imported.

How prices are shown

Prices belong to an exact model listing and gateway. Token rates are normalized to USD per million tokens, with context tiers and upstream-route conditions preserved where supplied. LiteLLM rates describe configured upstream providers, not a hosted LiteLLM service. Missing rates are shown as Not published in this snapshot; zero is shown only when explicitly published. Platform fees, credit fees, cache, media, tool charges and negotiated discounts can change your bill. Prices are reviewed snapshots, not live quotes.

How listings are maintained

Use, free and ranked collection filters use OpenRouter’s published metadata. Chinese Models is a curated cross-provider collection matched by documented family names, including named derivatives; it describes developer origin, not hosting location. Frontier and value shortlists use its Artificial Analysis intelligence ordering; fastest uses its recent median throughput ordering. Rankings are collected for review with each OpenRouter refresh. Exact criteria and a source link appear when selecting a collection. Listings without OpenRouter metadata are not included in its use, free or ranked filters.

We collect model IDs from official sources, review changes, and publish an approved snapshot. A failed refresh preserves the last successful listings. “No longer listed” means a model disappeared from repeated source snapshots; it does not establish retirement or lack of support. Compatibility registries require configuration and are not hosted inventories.

Source coverage was researched on 15 September 2026. Each imported listing has its own last-observed date. Send a correction.