We’ve switched Augment’s default model to GPT-5.4. It’s available in Intent, the Auggie CLI, and our IDE extensions. Existing threads won’t change unless you switch models.
Where GPT-5.4 stands out
GPT-5.4 performs especially well at multi-agent orchestration. In our testing it stays anchored to a plan longer, updates that plan when constraints change, and closes the loop with verification instead of restarting mid-run.
The difference is:
- fewer tool-state failures
- fewer “re-derive the approach” turns
- smoother coordination across multiple agents
These improvements show up most clearly in Intent workflows, where agents execute a spec end-to-end.
A strong everyday model
GPT-5.4 is also a reliable daily driver for everyday development work. The same improvements in reasoning and tool use help with tasks like quick edits, debugging, and general coding workflows.
More efficient and reliable on complex tasks
On harder tasks like complex refactors, multi-file reasoning, and architectural planning, GPT-5.4 uses ~18–20% fewer tokens in our internal testing. It also tracks workflow state more reliably and completes multi-step sequences without repeating work.
Good workflows to test:
- run a refactor
- ask for an architectural plan
- start an Intent workflow and let it orchestrate the work end-to-end