ByteHorizon
Unified AI gateway

One API. Every model your business needs to go global.

ByteHorizon routes your requests to the best AI models worldwide — OpenAI, Anthropic, and self-hosted open models — through a single OpenAI-compatible endpoint, billed in one place.

Browse models

# one endpoint, any model

client = OpenAI(

base_url="https://api.bytehorizonai.com/v1",

api_key="sk-...",

)

client.chat.completions.create(

model="bh-chat-pro",

messages=[...],

)

7

Models

3

Providers

200K

Max context

1

Unified endpoint

Drop-in OpenAI compatible

Point your existing SDK at our base URL. No new client, no rewrites.

One key, every provider

Route to OpenAI, Anthropic, and self-hosted open models behind a single endpoint.

Pay only for what you use

Per-token pricing with real-time spend tracking — no idle infrastructure costs.

Spin up keys in seconds

Issue scoped, budget-capped keys for every team, app, or customer you serve.

Models on the shelf

A sample of what's live right now — the full list is pulled from the gateway.

View all
B

ByteHorizon Chat Pro

ByteHorizon

Flagship general-purpose chat model. Strong reasoning and instruction-following for everyday production traffic.

32K ctx$2 / $8 per 1M
B

ByteHorizon Chat Lite

ByteHorizon

Lightweight and fast — built for high-volume, latency-sensitive chat and classification tasks.

32K ctx$0.6 / $2.4 per 1M
O

GPT-4o

OpenAI

OpenAI's natively multimodal flagship model, routed through the gateway with one unified key.

128K ctx$5 / $15 per 1M
A

Claude 3.5 Sonnet

Anthropic

Anthropic's balanced model for coding, writing, and analysis — strong instruction-following at mid-tier cost.

200K ctx$3 / $15 per 1M

How it works

01

Get an API key

Create a key from your dashboard — scoped, budgeted, revocable anytime.

02

Call one endpoint

POST to /v1/chat/completions with any model id from the shelf below.

03

Pay as you go

Usage and cost are tracked per call — see exactly what each model costs you.

Ready to ship?

Get a key, point your SDK at our endpoint, and start serving every model your product needs — without juggling a dozen vendor accounts.