Open-source models · HALO Docs

The open-source gateway routes listed canonical model IDs to validated custom inference providers while preserving an OpenAI-compatible chat response.

#Make a request

curl https://api.agihalo.com/qwen/v1/chat/completions \
  -H "Authorization: Bearer $HALO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3.6-35b-a3b",
    "messages": [{"role": "user", "content": "Reply with OK"}],
    "max_tokens": 16,
    "stream": false
  }'

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.HALO_API_KEY,
  baseURL: "https://api.agihalo.com/v1",
});

const response = await client.chat.completions.create({
  model: "qwen/qwen3.6-35b-a3b",
  messages: [{ role: "user", content: "Reply with OK" }],
});

This SDK base URL accepts configured OpenAI and official DeepSeek model IDs as well as listed open-model IDs. HALO selects exactly one runtime from the canonical model value.

#Runtime contract

Text-only messages and visible assistant output.
n and best_of, when present, must be 1.
Tools, function declarations, tool-role messages, audio, images, and multimodal content are rejected before provider dispatch.
Runtime normalizes to stream: false and caps output at 32,768 tokens.

#Model families

The production catalog groups listed models across Qwen, Llama, DeepSeek, Kimi, Mistral, Gemma, GLM, MiniMax, Nemotron, MiMo, GPT-OSS, and AllenAI families. Use the dashboard's Models page for the current project-visible IDs.

Family-bound base URLs are /qwen/v1,/llama/v1, /deepseek/v1,/kimi/v1, /mistral/v1,/gemma/v1, /glm/v1,/minimax/v1, /nemotron/v1,/mimo/v1, /gpt-oss/v1, and/allenai/v1. A family URL rejects a canonical model ID belonging to another family.

#Supplying inference

If you operate an OpenAI-compatible HTTPS endpoint, read the Keeper contract before registration. Model discovery, baseline checks, bounded health probes, deterministic metering, and payout rules apply.