According to a recent LinkedIn post from Together AI, the company is emphasizing the growing practicality of open models as a foundation for agentic coding workflows. The post outlines a five-layer architecture for this approach, described as model, inference, gateway or router, harness, and tools, which Together AI refers to collectively as the MIGHT stack.
The post suggests that each layer in the MIGHT stack can be selected independently, enabling developers to combine different open models with various infrastructure and orchestration providers. Examples cited include pairing models such as GLM-5.3 Flash or Kimi K3 with providers like Together AI, harnesses such as OpenCode, and developers’ own skills and MCP servers.
As shared in the LinkedIn post, this modular architecture is presented as a way to allow coding stacks to evolve as new models and tools emerge. For investors, this focus on flexible, open-model infrastructure may indicate Together AI’s intent to position itself as a key provider in the expanding ecosystem of agentic coding, potentially driving usage-based revenue as developers adopt and upgrade their stacks.
The post also notes that Together AI provides deeper technical guidance on how each layer functions and how to build an open model stack. This emphasis on developer education and open tooling may strengthen the company’s relationships with software teams, enhance platform stickiness, and support long-term relevance in the competitive AI infrastructure market.