Model catalog entries

11

11 enabled in the live catalog

Provider families

4

Live availability is keyed by family, not by a provider-native endpoint

Tier entitlements

0

Select a tenant to see live entitlements

Provider credentials

0

Select a tenant to inspect BYOK status

Catalog

Live model catalog

Treat this as the publish list. If a model does not appear here, the gateway should not advertise it.

cheap/gpt-5.4

Pooled OpenAI-backed reasoning and general chat lane

enabledpublic
cheapopenaipooled-accountstreamingtoolsjsonno browser fallback
Route policy: shared-premiumRouting key: pooled:openai:gpt-5.4Meter: cheap_tokensMax context: 128,000Legacy aliases: cheap-chat, cheap-reasoning
cheap/gpt-5.4-coder

Pooled coding lane with stronger stickiness requirements

enabledpublic
cheapopenaipooled-accountstreamingtoolsjsonno browser fallback
Route policy: shared-premiumRouting key: pooled:openai:gpt-5.4-coderMeter: cheap_tokensMax context: 128,000Legacy aliases: cheap-coder
cheap/gpt-5.4-long

Pooled long-context lane with async-friendly queueing

enabledpublic
cheapopenaipooled-accountstreamingno-toolsjsonbrowser fallback
Route policy: bulk-priorityRouting key: pooled:openai:gpt-5.4-longMeter: cheap_tokensMax context: 256,000Legacy aliases: cheap-longctx
resale/gpt-5.4

Official OpenAI resale lane behind the universal OpenAI-compatible API

enabledpublic
resaleopenaiofficial-apistreamingtoolsjsonno browser fallback
Route policy: provider-resaleRouting key: official:openai:gpt-5.4Meter: resale_openai_tokensMax context: 128,000Legacy aliases: none
resale/claude-sonnet

Official Anthropic resale lane behind the universal OpenAI-compatible API

enabledpublic
resaleanthropicofficial-apistreamingtoolsjsonno browser fallback
Route policy: provider-resaleRouting key: official:anthropic:claude-sonnetMeter: resale_anthropic_tokensMax context: 200,000Legacy aliases: none
resale/gemini-pro

Official Gemini resale lane behind the universal OpenAI-compatible API

enabledpublic
resalegeminiofficial-apistreamingtoolsjsonno browser fallback
Route policy: provider-resaleRouting key: official:gemini:gemini-proMeter: resale_gemini_tokensMax context: 1,000,000Legacy aliases: none
resale/glm-4.5

Official GLM resale lane behind the universal OpenAI-compatible API

enabledpublic
resaleglmofficial-apistreamingtoolsjsonno browser fallback
Route policy: provider-resaleRouting key: official:glm:glm-4.5Meter: resale_glm_tokensMax context: 128,000Legacy aliases: none
byok/openai/gpt-5.4

Tenant-owned OpenAI passthrough lane

enabledpublic
byokopenaitenant-credentialstreamingtoolsjsonno browser fallback
Route policy: provider-byokRouting key: byok:openai:gpt-5.4Meter: byok_openai_tokensMax context: 128,000Legacy aliases: none
byok/anthropic/claude-sonnet

Tenant-owned Anthropic passthrough lane

enabledpublic
byokanthropictenant-credentialstreamingtoolsjsonno browser fallback
Route policy: provider-byokRouting key: byok:anthropic:claude-sonnetMeter: byok_anthropic_tokensMax context: 200,000Legacy aliases: none
byok/gemini/gemini-pro

Tenant-owned Gemini passthrough lane

enabledpublic
byokgeminitenant-credentialstreamingtoolsjsonno browser fallback
Route policy: provider-byokRouting key: byok:gemini:gemini-proMeter: byok_gemini_tokensMax context: 1,000,000Legacy aliases: none
byok/glm/glm-4.5

Tenant-owned GLM passthrough lane

enabledpublic
byokglmtenant-credentialstreamingtoolsjsonno browser fallback
Route policy: provider-byokRouting key: byok:glm:glm-4.5Meter: byok_glm_tokensMax context: 128,000Legacy aliases: none

Operator reference

Launch inspector

Compare tiers, entitlements, credentials, and pooled caveats against the live catalog.

cheap/*
Pooled inference with strict scrub and privacy caveats
Shared account routing
resale/*
Official-provider resale behind the same OpenAI-compatible API
Provider-backed
byok/*
Tenant-owned provider credentials, masked and logged by the gateway
Tenant scoped
Plan signal
standard
Portal defaults until live entitlement payload loads
cheap/*

Pooled inference is available in the default portal contract.

enabled
OpenAI-compatible /v1/*Sticky thread affinityStrict scrub mode

Models: cheap/gpt-5.4, cheap/claude-sonnet, cheap/longctx

resale/*

Official-provider pass-through is available to this tenant.

enabled
Provider-native cost basisOpenAI-compatible APINo pooled fallback

Models: resale/gpt-5.4, resale/claude-sonnet

byok/*

BYOK is visible in the portal contract and becomes active once tenant provider credentials are configured.

needs-credentials
Tenant provider credentialsKey maskingUsage logging

Models: byok/openai/gpt-5.4, byok/anthropic/claude-sonnet

Provider credential status
Backend credential rows have not been loaded yet, so the portal falls back to the scaffold.
Fallback
OpenAI provider credential slot

Add a tenant-owned OpenAI API key for byok/openai/* routes.

placeholder
Display labelAPI keyOptional project label

Portal fallback: the live credential rows are shown in the dashboard when the backend surface is reachable.

Anthropic provider credential slot

Store an Anthropic key for byok/anthropic/* routes.

placeholder
Display labelAPI keyOptional workspace name

Keep this isolated from pooled tiers; no fallback to cheap/*.

Gemini provider credential slot

Reserved for future tenant-managed Gemini routing.

not-configured
Display labelAPI keyProject identifier

Intentionally scaffolded in the portal before the backend adapter lands.

GLM / Z.AI provider credential slot

Reserved for future tenant-managed GLM routing.

not-configured
Display labelAPI keyRegion notes

This section is visible so the UX contract is ready before the backend integration.

criticalPooled mode keeps strict scrub on by default

Cheap/* traffic is routed through pooled accounts. Keep secrets, tokens, and personal data out of those requests unless you explicitly accept the pooled-mode caveats.

warnHistory cleanup is an operator workflow, not a guarantee

The portal exposes the cleanup story as an operator workflow, but pooled requests can still be visible in the shared account history. Treat them as non-dedicated traffic.

infoBYOK data stays tenant-scoped

BYOK routes should never use pooled fallbacks. The live portal keeps tenant provider credentials masked and tenant-scoped.