AI Models
OpsTower supports a range of AI models from multiple providers. You can configure default models for different agent types and override them on a per-agent basis.
Available Models
Section titled “Available Models”Anthropic
Section titled “Anthropic”| Model | Capability | Tier |
|---|---|---|
| Claude Opus 4 | Most capable | Premium+ |
| Claude Sonnet 5 | Balanced | Standard (system default) |
| Claude Haiku 4.5 | Fastest | Standard |
OpenAI
Section titled “OpenAI”| Model | Capability | Tier |
|---|---|---|
| GPT-5.5 | Most capable | Premium |
| GPT-5.6 Sol | Frontier | Premium |
| GPT-5.6 Terra | Balanced | Premium |
| GPT-5.6 Luna | Fast, cost-effective | Standard |
| Model | Capability | Tier |
|---|---|---|
| Gemini 3.1 Pro | Most capable | Premium |
| Gemini 3.5 Flash | Balanced | Standard |
| Gemini 3.1 Flash Lite | Lightweight | Standard |
Moonshot AI
Section titled “Moonshot AI”| Model | Capability | Tier |
|---|---|---|
| Kimi K3 | Frontier | Premium |
- Kimi K3 is Moonshot AI’s frontier model with native visual understanding and a 1M-token context window. It always runs with thinking mode enabled (with configurable reasoning effort) and is served via Moonshot’s hosted API.
Cloudflare Workers AI
Section titled “Cloudflare Workers AI”| Model | Capability | Tier |
|---|---|---|
| MiniMax M3 | Open-source, agentic | Standard |
| GLM-5.2 | Open-source, agentic | Standard |
| Kimi K2.7 Code | Open-source, agentic | Standard |
These open-source models run on Cloudflare Workers AI via the env.AI binding.
- MiniMax M3 is an open-source frontier model with a 1M-token context window, strong at coding and agentic tasks.
- GLM-5.2 is Z.ai’s open-source frontier model with a 262K-token context window, built for agentic coding, long-codebase reasoning, multi-step planning, and tool-augmented agents.
- Kimi K2.7 Code is Moonshot AI’s open-source coding model with a 262K-token context window, featuring multi-turn tool calling, configurable thinking mode, vision inputs, and structured outputs.
Model Tiers
Section titled “Model Tiers”Models are grouped into three tiers that correspond to plan levels:
- Standard – available on all plans (Explore, Operate, Scale)
- Premium – available on Operate (Pro) and Scale (Business) plans
- Premium+ – available on Scale (Business) only
Web Search and Fetch Tools
Section titled “Web Search and Fetch Tools”The tools available for web access vary by provider:
| Provider | Web Tools |
|---|---|
| Anthropic | Web Search, Web Fetch |
| Google Search, URL Context | |
| OpenAI | Web Search |
| Moonshot AI | Web Fetch (+ browser tools) |
| Workers AI | Web Fetch (+ browser tools) |
Moonshot AI (Kimi K3) and Workers AI models (MiniMax M3, GLM-5.2, Kimi K2.7 Code) have no provider-native web search. They read the web via a generic fetch tool and the built-in browser tools (browser_execute), which are available regardless of provider.
Setting Default Models
Section titled “Setting Default Models”You can configure a default model for each agent scope on the Settings page. The available scopes are:
- Debugger agents – the default model used by all Debugger agents
- Analyst agents – the default model used by all Analyst agents
- PostHog reports – the model used when generating reports from PostHog connections
- Amplitude reports – the model used when generating reports from Amplitude connections
- Mixpanel reports – the model used when generating reports from Mixpanel connections
To set a default:
- Navigate to the Settings page.
- Find the model scope you want to configure.
- Select a model from the dropdown.
- Save your changes.
Per-Agent Override
Section titled “Per-Agent Override”You can override the default model for any individual agent:
- Open the agent you want to configure.
- Go to the agent’s General settings.
- Select a different model from the model dropdown.
- Save.
This agent will now use the selected model regardless of the scope default.
Model Resolution Order
Section titled “Model Resolution Order”When OpsTower determines which model to use for an agent, it follows this resolution order:
- Per-agent override – if a model is explicitly set on the agent, it is used.
- Scope default – if no per-agent override exists, the default for the agent’s scope (e.g., Debugger, Analyst) is used.
- System default – if no scope default is configured, the system default (Claude Sonnet 5) is used.