Skip to main content

​
Current models

Augment uses world-class large language models together with our Context Engine. We currently support the following models:
  • Claude Fable 5 by Anthropic
  • Claude Haiku 4.5 by Anthropic
  • Claude Opus 4.5 by Anthropic
  • Claude Opus 4.6 by Anthropic
  • Claude Opus 4.7 by Anthropic
  • Claude Sonnet 4.5 by Anthropic
  • Claude Sonnet 4.6 by Anthropic
  • Gemini 3.1 Pro by Google
  • GPT-5.1 by OpenAI
  • GPT-5.2 by OpenAI
  • GPT-5.4 by OpenAI
  • GPT-5.5 by OpenAI
  • Kimi K2.6 by Moonshot AI (hosted on Fireworks and Baseten)
  • Prism (Claude + Gemini)
  • Prism (GPT + Kimi)

​
Prism routing options

Prism lets Augment choose the best-fit model for each request. Instead of locking you into a single model, each Prism option routes within a curated model family based on the task, context, and current system conditions.
  • Prism (Claude + Gemini) routes between Claude Opus 4.7, Claude Sonnet 4.6, and Gemini 3.0 Flash.
  • Prism (GPT + Kimi) routes between GPT-5.5, GPT-5.4, and Kimi K2.6.

​
Choosing a model

  • In Cosmos, choose the model when configuring an expert or starting a session.
  • In Auggie CLI, use the /model slash command or pass the --model flag with the desired model.
  • Your selection can be changed at any time.
If you don’t pick a model, Augment will use your last selection or the default set by your organization. For how models are billed and current per-model rates, see Token-Based Pricing.

​
Feature compatibility

All listed models support the core Augment features:
  • Deep code understanding with Augment’s Context Engine
  • Tool use (integrations and MCP), file edits, and multi-step planning
Some behaviors (e.g., wording or style) may vary slightly between models. We’ll continue to refine prompts and guardrails to provide a consistent developer experience.

​
Notes and transparency

  • We may roll out model updates gradually. If you’re part of a staged rollout, different workspaces or teammates may see updates at different times.
  • For troubleshooting, you can share the request ID with our support team; they can confirm which model handled a specific request.
If you have questions about model availability or want to participate in early access programs, please reach out via Support.