Introducing Together Link: Run open models inside the harness you already use →

The Open Source AI Stack: models, inference, routers, harnesses & tools →

📈 GLM-5.3 vs. GLM 5.3 Flash on DeepSWE →

🚀 DeepSeek V4 Pro 0813 vs. GPT-5.6 Sol on DeepSWE →

⚡ On-demand B200s now available on Together GPU Clusters →

All blog posts
Inference

Published 10/5/2026

Together Link: frontier-quality open models in the harness you already use

Frontier open models in the coding agent you already use, for over 50% less

Summary

Together Link connects the harness your team already uses to the best open models on Together AI and cuts your spend by over 50%. Same agent, same workflow, a fraction of the bill.

Coding agents are now commonplace for engineering teams, and every task you give them, from a one-line fix to a full rewrite, often runs on the same premium model. At scale that gets expensive quickly: engineering orgs are spending anywhere from tens of thousands to millions of dollars a month on closed models.

Open models have meaningfully closed the gap between top models from closed labs. Leaders like Kimi K3 and GLM 5.3 now handle many of the hardest coding tasks at a fraction of the per-token price, while small and capable models like GLM 5.3 Flash and DeepSeek V4.1 Flash handle everyday coding tasks. Together Link brings them into the tools your team already knows.

Works where you already work

Together Link supports Claude Code, Claude Desktop, Codex (in the ChatGPT app and the CLI), OpenCode, and Pi. Your team keeps working normally: settings and logins stay as they were, there's nothing new to learn, and going back to the native closed models takes one command. You can be up and running on open models in just a few minutes.

Getting started

curl -fsSL https://link.together.ai/install | bash
  1. Run one command to set up Together Link: Your agent opens exactly as before, now connected to Together AI with your Together AI API key. Create an account for an API key.
  2. Use “Auto” to pick a model with the Together AI Router: It reads each session's first task and sends it to the right model: quick fixes go to fast, low-cost models, and hard problems get frontier capability. If you bring an Anthropic key, it routes between Opus 5.5 and GLM 5.3. Without one, it routes between GLM 5.3 and GLM 5.3 Flash. Routing happens once per session, so prompt caching keeps working.
  3. See your savings after each session. A per-session tracker shows what you spent next to what the same session would have cost on Opus 5.5.

Billing runs on your existing Together API key, against serverless pay-as-you-go or credit packs. You don't need a separate contract.

Built on the serverless platform developers already choose

Together Link runs on Together's serverless inference, the same infrastructure developers already pick for these models on OpenRouter. Together AI serves the largest share of OpenRouter tokens for DeepSeek V4.1 Flash (40.8%), GLM 5.3 Flash (28.2%), and Kimi K3 (23.1%), with competitive speed and pricing across the board. (As of 9/30/2026 footnote)

Next steps

Create a Together AI account.

Read our docs.

Rolling it out across your engineering org? Talk to our team and we'll help you plan it.

‍

Related articles

View All
View All
Inference
Together Link: frontier-quality open models in the harness you already use
Inference
Canary rollouts: upgrade models in production without downtime
Inference
How a global fintech scaled coding agent traffic with Dedicated Model Inference
Inference
Migrating from closed to open source models, Together
Inference
The Open Source AI Stack
Inference
A/B test models in production
Inference
Autoscaling endpoints for LLM inference
Inference
Configuring Dedicated Model Inference
Inference
The production platform for open-weight AI inference
Inference
What does 99.9% uptime mean for inference?
Inference
Together AI brings Thinking Machines Lab’s new model Inkling on day 0
Inference
Open, convenient and predictable: Introducing Provisioned Throughput
Inference
Kimi K2.7 Code vs Claude Fable 5: Landing pages that cost 94% less
Inference
Serving MiniMax-M3 for efficient inference: Unlocking 1M-Token Context and Multimodality Without Regrets
Inference
How Together AI built the world’s fastest speech-to-text stack
Inference
Benchmarking inference at scale: coding agents
Text reads: Inference in production: coding agent benchmarks with together.ai logo on pastel gradient background.
Inference
Together AI and Pearl Research Labs Team Up to Reduce the Cost of AI Inference
Together AI and Pearl Research Labs team up to reduce the cost of AI inference on a soft gradient background.
Inference
Serving DeepSeek-V4: why million-token context is an inference systems problem
Together.ai banner with title: Serving DeepSeek-V4, why million-token context is an inference problem.
Inference
Deploy and inference any model from HuggingFace
Together.ai logo with text: Deploy and inference any model from HuggingFace on a pastel gradient background.
Inference
Foundational research powering efficient inference at scale
Text stating foundational research powering efficient inference at scale with together.ai logo on gradient background.
Inference
Introducing Dedicated Container Inference: Delivering 2.6x faster inference for custom AI models
Table listing Avatar versions 2.4, 2.3, 2.2 with statuses Ready, Running, and Ready.
Inference
Optimizing inference speed and costs: Lessons learned from large-scale deployments
Graph showing trade-off between cost efficiency and inference speed with Pareto frontier and labeled providers.
Inference
Learn how Cursor partnered with Together AI to deliver real-time, low-latency inference at scale
Inference
Together AI delivers fastest inference for the top open-source models
Bar chart comparing output speed of Together AI, Amazon, and Google Vertex with Together AI fastest.
Inference
Dynamic AI agent testing for the real world with Collinear Simulations and Together Evals
Text about dynamic AI agent testing using Together AI Evals and Collinear Simulations with their logos.
Inference
Improved Batch Inference API: Enhanced UI, Expanded Model Support, and 3000× Rate Limit Increase
Text saying together Batch Inference API and a description of running massive inference jobs with 30B tokens.
Inference
Together Evaluations: Benchmark Models for Your Tasks
Together.ai homepage heading with text: Benchmark models for your tasks and a blue Learn More button.
Inference
Together AI Delivers Top Speeds for DeepSeek-R1-0528 Inference on NVIDIA Blackwell
Bar chart showing output speed in tokens/second of AI providers, ranging from Together.ai 355 to DeepSeek 23.
Inference
Introducing the Together AI Batch API: Process Thousands of LLM Requests at 50% Lower Cost
Inference
Introducing Together Code Sandbox & Together Code Interpreter: SOTA code execution for AI
Text on dark background: Introducing Together Code Sandbox Together Code Interpreter.
Inference
Together Code Sandbox: the most robust infrastructure for building AI coding products at scale
Text 'Together Code Sandbox' with a diagram showing Agent and LLM connecting to Code Sandbox.
Inference
Together Code Interpreter: execute LLM-generated code seamlessly with a simple API call
Text 'Together Code Interpreter' and diagram showing LLM generates code, TCI executes code, agent uses output.
Inference
Introducing Together Chat: use DeepSeek R1 for free, hosted in North America
Together.ai interface showing DeepSeek R1 chat answering a quantum physics question simply about particles.
Inference
Deploy Leading AI Models Accelerated by NVIDIA NIM on Together AI
Text says Deploy NVIDIA NIM Microservices to Together AI with a hexagonal icon of connected green nodes.
Inference
On-demand dedicated endpoints: run inference with unmatched price-performance & control at scale
Text about dedicated endpoints with an illustration of increasing blue cube stacks connected to a dial.
Inference
Deploy DeepSeek-R1 at scale: Fast, secure serverless APIs and large-scale Together Reasoning Clusters
Text promoting DeepSeek-R1 deployment with fast, secure serverless APIs and large-scale Together Reasoning Clusters.
Inference
Generate images with specific styles using Flux LoRAs on Together AI
Text about generating images with Flux LoRAs with pixel art, 3D, and line art llama styles and together.ai logo.
Inference
Announcing Together Inference Engine 2.0 with new Turbo and Lite endpoints
Inference
Using Axiomic to build multi agent chat with Together API
Inference
Announcing v1 of our Python SDK
Inference
Announcing function calling and JSON mode
Inference
Building your own RAG application using Together AI and MongoDB Atlas
Inference
Building your own RAG application using Together AI and Langchain
Inference
Building your own RAG application using Together AI and LlamaIndex
Inference
Introducing the Together Embeddings endpoint — Higher accuracy, longer context, and lower cost
Inference
Filter responses of any model with Llama Guard or your own safety model
Inference
Announcing Together Inference Engine – the fastest inference available
Inference
Announcing Together Custom Models. Build a state-of-the-art LLM with Together AI — and own the model.
Inference
Together AI launches full stack for developers to build with open-source AI
Inference
RedPajama 7B now available, instruct model outperforms all open 7B models on HELM benchmarks
Inference
RedPajama-INCITE-3B, an LLM for everyone
Inference
Releasing 3B and 7B RedPajama-INCITE family of models including base, instruction-tuned & chat models
Inference
RedPajama, a project to create leading open-source models, starts by reproducing LLaMA training dataset of over 1.2 trillion tokens
Inference
OpenChatKit now runs on consumer GPUs with a new 7B parameter model
Inference
Announcing OpenChatKit
Inference
How to Build a State-of-the-Art Search Stack for LLMs: RAG, Reranking, and Reinforcement Learning
Text on dark background: How to Build a SOTA Search Stack for LLMs: RAG, Reranking, and RL
Inference
Best practices to accelerate inference for large-scale production workloads
Inference
Together Link: frontier-quality open models in the harness you already use
Fine-Tuning
How to train your own Jev for $17
Inference
Canary rollouts: upgrade models in production without downtime
Inference
How a global fintech scaled coding agent traffic with Dedicated Model Inference
Inference
Migrating from closed to open source models, Together
Fine-Tuning
Together AI expands fine-tuning service with more models, live metrics, and finer controls
GPU Clusters
Introducing preemptible compute: the same compute, half the price
Research
To Infinity and Beyond: ThunderKittens Now on NVIDIA Vera Rubin NVL72!
Inference
The Open Source AI Stack
Model Library
GLM-5.3 vs. GLM-5.3 Flash on DeepSWE: Cost, Coding, and Routing
Model Library
GLM-5.3 vs. GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing
Model Library
GLM-5.3 vs. Claude Fable 5 on DeepSWE: Cost, Coding, and Routing