As part of our commitment to supporting the developer community with reliable, scalable Thai AI infrastructure, Typhoon launched Typhoon 2 API Pro earlier this year through Together AI. This article originally introduced that offering and is now updated to reflect the current status of API Pro, including upcoming changes, deprecation timelines, and recommended migration paths.
With this update, we want to ensure developers have full clarity on how API Pro is evolving and what to expect next.
🚨 Important Update: Typhoon API Pro Deprecation
The original Typhoon API Pro (Typhoon 2.1 Gemma on Together AI) will be sunsetting on 31 December 2025.
This decision allows us to prepare for a much stronger production environment. In early 2026, we will launch a new, next-generation Typhoon API on Amazon Web Services (AWS) featuring:
-
Typhoon 2.5 (the more recent text model)
-
Typhoon OCR 1.5
-
Typhoon ASR Real-Time
We expect availability in Q1 2026, and will announce exact rollout dates soon.
If You Currently Use Typhoon API Pro
While the Together AI–based API Pro ends on 31 December 2025, you still have multiple supported options:
1) Continue using Typhoon via our free hosted API
You can access Typhoon models, including Typhoon 2.5, through our free API offered here at opentyphoon.ai
Requests are rate-limited and may not be suitable for high-throughput production workloads. See this guide.
2) If you require high-volume or production workloads today
We recommend migrating to one of the following:
-
Float16’s LLM-as-a-Service as they've already hosted and offered Typhoon models including Typhoon 2.1 Gemma.
-
Provision your own GPU/ VM (via e.g. Float16, Runpod, AWS, Azure, GCP) and host Typhoon models in your own environment. See the hardware guideline here.
These options ensure continuity while we prepare our AWS-native production API.
📦 Original API Pro Release Details (Historical)
Below is the original feature overview from the Together AI launch, kept here for historical reference and for users who need to understand the prior API:
What Was New (June 2025 Update)
Typhoon 2.1 Gemma available on Together AI
Typhoon 2.1 Gemma became the primary production-grade model for API Pro, offering:
-
Lower latency
-
Higher throughput
-
Seamless serverless scaling
Endpoint: scb10x/scb10x-typhoon-2-1-gemma3-12b
Deprecation of Older Models
-
Typhoon-2-8B was removed earlier in 2025
-
Typhoon-2-70B was scheduled for deprecation on August 20, 2025
-
Typhoon 2.1 Gemma became the recommended upgrade path
These changes are now superseded by the full deprecation of the Together-based API Pro at the end of 2025.
Key Features (Historical)
-
Leverage Together AI’s serverless inference infrastructure to deliver fast responses while minimizing costs through per-token billing
-
Compatible with popular frameworks and easy to integrate with existing AI pipelines
-
Built for production-grade Enterprise AI applications, offering reliability, uptime, and scalability
While this infrastructure will soon sunset, we remain committed to delivering an even better production environment on AWS.
Pricing (Historical)
The original API Pro pricing on Together AI:
-
Typhoon 2.1 Gemma 12B: $0.20 per 1M tokens
-
Typhoon 2 Instruct 70B: $0.88 per 1M tokens
These prices no longer apply after the deprecation date.
Quickstart Guide (Historical Documentation)
This section remains unchanged for developers who want to reference earlier integrations or maintain legacy workflows until 31 December 2025.
Try out Typhoon Models in Together’s playground
Register for a Together.ai account. New accounts get $1 free credit to help you get started.
You can test the Typhoon models on the Together Playground.
Get API Access
- Set your API key as an environment variable named
TOGETHER_API_KEY
- Make your first API call
Using Typhoon API Pro with Official Together Libraries
Together AI provides official libraries for Python and TypeScript. You can install them using pip install together for Python and npm install together-ai for Typescript.
Send your first Typhoon 2 API Pro request and start generating a response:
Python
Typescript
OpenAI Compatibility
Together API endpoints are fully compatible with the OpenAI API. If your application is already using OpenAI’s client libraries, you can configure it to point to Together API servers and start using Typhoon models.
Using Together APIs with OpenAI Client Libraries
- Set the
api_keyto your Together API key. You can find your API key in your Settings page. - Update the
base_urltohttps://api.together.xyz/v1
Python
Typescript
Join the Typhoon Community
We’re grateful to everyone who adopted Typhoon API Pro during its time on Together AI. Your feedback and support directly informed our decision to build a stronger, more scalable, AWS-native Typhoon API launching in 2026.
Join our community to stay updated on release dates, migration guides, and new capabilities: https://discord.gg/9F6nrFXyNt
🚀 What’s Next
We are actively preparing:
-
A fully managed production API on AWS
-
New capabilities across text, audio, OCR, and agentic workflows
-
Lower latency, higher reliability, and improved developer experience
-
Migration guides for users transitioning from Together AI or self-hosted environments
You can continue using our free hosted API at opentyphoon.ai while awaiting the new release.
For full Typhoon documentation, visit https://docs.opentyphoon.ai/
