Skip to content
Typhoon API Docs

Typhoon Documentation & Developer Guide

👋 Welcome to Typhoon Documentation.

Overview: Where & How to Access Typhoon Models

Typhoon is available on multiple platforms. Choose what suits your workflow:

0. Web Playground 🌐

  • Best for: Fast experimentation in your browser
  • No setup required — test Typhoon models with just a prompt.
  • Available in text and multimodal capabilities:

— 🔗 Playground for Typhoon 2

— 🔗 Playground for Typhoon T1 (research preview)

— 🔗 Playground for Typhoon Audio (research preview)

— 🔗 Playground for Typhoon Vision (research preview)

1. API Access 🔌

Best for: App developers and production deployments

Typhoon offers two hosted APIs to suit your needs:

1.1 Typhoon API ✅ (Free Tier)

Key Features
  • OpenAI Compatible API: Easy integration with existing OpenAI-based applications
  • Streaming Support: Real-time token streaming for interactive applications
  • Hosted by Typhoon — ideal for prototyping and light usage
  • Rate limits apply
Available Models
Model IDSizeDescriptionContext WindowRate LimitsRelease Date
typhoon-v2-8b-instruct8BImproved model for general Thai language tasks8K tokens5 req/s, 50 req/m2024-12-19
typhoon-v2-70b-instruct70BOur most powerful model with advanced Thai language capabilities8K tokens5 req/s, 50 req/m2024-12-19
typhoon-v2-r1-70b-preview70BThai reasoning model based on DeepSeek R1 70B Distill and Typhoon2 70B Instruct + SFT merged8K tokens5 req/s, 50 req/m2025-03-17

📘 See the navigation bar to explore Typhoon API Documentation

1.2 API via Together.ai ⚡ (Scalable + Cost Effective)

We offer Typhoon API Pro with a collaboration with Together AI, a powerful cloud platform that allows you to run and customize AI models faster and easier.

Best for: Production workloads requiring higher reliability and throughput

Pay-as-you-go pricing with higher rate limit: Learn more

Available Models
Get Started

📘 Read the quickstart guide

2. Cloud Marketplace Access ☁️

Typhoon models are currently making the way to be available on:

  • Azure Model Catalog (Typhoon 1.5 available, newer models are coming soon)
  • AWS Marketplace (coming soon)

Please refer to respective marketplace listings for setup instructions.

3. Run Locally 💻

These tools support local use:

3.1 🤗 Hugging Face

Run Typhoon models locally, fine-tune them, or integrate into custom ML pipelines using the popular Hugging Face ecosystem.

Key Features
  • Transformers Integration: Easily plug into the Hugging Face transformers library
  • Offline & On-Premise Ready: Ideal for teams with privacy or infrastructure requirements
  • Community Tools & Datasets: Leverage the broader Hugging Face ecosystem for research or experimentation
Model familyAvailable SizesDescription
Typhoon2-Text1b, 3b, 7b, 8b, 70bLatest Official Text LLMs
Typhoon2-R170bReasoning model based on DeepSeek R1 70B Distill & Typhoon2 70B Instruct
Typhoon-T13bThai reasoning model built on top of Typhoon 2 3B Instruct
Typhoon2-Audio8bEnd-to-end, speech-to-speech model
Typhoon2-Vision7bVision-language model supporting both image and video inputs

🔗 Explore our Hugging Face collections for the full list

3.2 Alternative Platforms

Key Features
  • Local Inference: Run models without relying on internet or external APIs
  • Cross-Platform Support: Works on Windows, macOS, and Linux
  • No Rate Limits: Unlimited usage, limited only by your hardware
  • Private & Customizable: Full control over deployment and environment

For example, Typhoon models are available with these platforms:

  • Ollama

    A simple command-line tool to run and manage open models locally. Lightweight, developer-friendly

    🔗 Typhoon on Ollama

    📘 Read a quickstart tutorial here

  • LM Studio

    A GUI-based desktop app for running local LLMs without code. Easy drag-and-drop interface. Supports model imports in GGUF format.

Get Started

  1. Sign up for an API key
  2. Check out the Quick Start Guide
  3. Explore the Examples for ideas on how to use the API

Support and Community