Available Models

All models accessible through a single OpenAI-compatible endpoint. Pricing is per 1M tokens (USD). Data is sourced live from the model catalog — automatically reflects current offerings and pricing.

List Models API

You can retrieve the model list programmatically:

GET /v1/models
Authorization: Bearer sk-bc-YOUR_API_KEY

Or without authentication for public pricing pages:

GET /api/v1/gateway/models/public

Google

Model IDDisplay NameInput / 1M tokensOutput / 1M tokensContextDescription
gemini-3.5-flashGemini 3.5 Flash$1.50$9.001,048,576Google's balanced multimodal model.
gemini-3.5-flash-liteGemini 3.5 Flash Lite$0.30$2.501,048,576Fast, low-cost Google model for high-volume workloads.

Moonshot

Model IDDisplay NameInput / 1M tokensOutput / 1M tokensContextDescription
kimi-3Kimi K3$0.60$2.50262,144Moonshot's Kimi K3 reasoning model.

OpenAI

Model IDDisplay NameInput / 1M tokensOutput / 1M tokensContextDescription
gpt-oss-120bGPT OSS 120B$0.35$1.40131,072OpenAI's open-weight 120B model.
gpt-oss-20bGPT OSS 20B$0.10$0.40131,072OpenAI's open-weight 20B model — fast and cheap.

MiniMax

Model IDDisplay NameInput / 1M tokensOutput / 1M tokensContextDescription
minimax-m3MiniMax M3$0.30$1.201,000,000MiniMax M3 general-purpose model.

Zhipu

Model IDDisplay NameInput / 1M tokensOutput / 1M tokensContextDescription
glm-5.2GLM 5.2$0.30$1.20131,072Zhipu GLM 5.2 multilingual model.

Model Aliases

Instead of specifying a model directly, you can use smart aliases:

AliasBehaviour
autoRoutes to the highest-priority available model on your plan (availability-first)
auto-cheapRoutes to the cheapest eligible model with available capacity (cost-first)

Bring Your Own Keys (BYOK)

If you need models not in the catalog, or want to use your own provider API keys, you can register Managed API Keys (BYOK) from the API Keys dashboard. Requests through Managed Keys are routed directly to your provider at your cost — billed at $0 by BroadComms, with full usage analytics.

Automatic Failover

The gateway routes each request across one or more inference providers for availability and cost optimisation. If a provider returns a 5xx error or times out, the request is automatically retried with a fallback provider. This happens transparently — you receive a successful response from the fallback with no changes required to your code.