Understanding models (E1/E2/E3 & Maxx)
What E1, E2, and E3 are
Emergent gives you three agents, E-1, E-2, and E-3, each running a different workflow. The LLM powering those agents is a separate selector. Think of the agent choice as picking how work gets done, not just how hard the model thinks.
- E-1: The default agent. Executes a clear step-by-step build workflow. Best for everyday features, bug fixes, and iterative changes.
- E-2: An integration-proving agent. Goes further to validate that the pieces it builds actually fit together, making it well-suited for moderate-complexity work involving multiple services or components.
- E-3: The autonomous agent. Runs E-1 as a sub-agent under its own orchestration loop, handling ambitious, open-ended goals with minimal hand-holding.
Start with E-1 or E-2
E-1 covers most everyday tasks efficiently. Reach for E-2 when you need integration confidence, and reserve E-3 for genuinely large or ambiguous goals.
When to use which agent
| Agent | Workflow style | Best for |
|---|---|---|
| E-1 | Step-by-step builder | Quick fixes, new components, standard integrations |
| E-2 | Integration-proving | Multi-service features, moderate refactors, reliability-sensitive changes |
| E-3 | Autonomous (runs E-1 as sub-agent) | Large architectural goals, open-ended projects, complex multi-step builds |
You select the agent before submitting your first message in the job creation form. The model you want to use is chosen separately via the model selector.
Agents cannot be switched mid-conversation
Once a job has started, the agent is fixed for that conversation. Even forking the conversation keeps the same agent. If you want a different agent, start a new job.
Maxx mode
Maxx mode is a deeper-reasoning toggle available in the prompt interface. When enabled, the agent applies more thorough reasoning to your request, useful when a problem genuinely warrants it.
Enabling Maxx consumes significantly more credits than a standard run. There is no published credit multiplier.
Maxx is a Pro-only feature
Maxx mode is available exclusively on the Pro plan. It is not accessible on Free or Standard plans.
When to enable Maxx
Identify the need
Ask yourself: Is the agent's default reasoning clearly insufficient for this problem? Maxx is most valuable for subtle, high-stakes logic where deeper thinking produces meaningfully better results.
Enable before submitting
Toggle Maxx mode in the prompt interface before you start the job. It applies to that job's reasoning depth.
Monitor credit usage
Check your credit balance after the job completes. Because Maxx consumes significantly more credits, consider breaking very large prompts into smaller sequential jobs if cost is a concern.
Choosing the right combination
A quick decision guide:
- Small, obvious change? → E-1, Maxx off.
- Standard feature or multi-service integration? → E-2, Maxx off.
- Large, open-ended, or highly autonomous goal? → E-3, Maxx off (add Maxx only if you're on Pro and the reasoning depth matters).
- On Pro and hitting a genuinely hard reasoning problem? → Enable Maxx on whichever agent fits the workflow.
Remember: if a job doesn't produce what you need, start a new job with a refined prompt or a different agent, you cannot switch agents within an existing conversation.
Iterate cheaply first
Start with E-1 or E-2 and Maxx off. A tighter follow-up in the same conversation, or a fresh job with a more specific prompt, is often more credit-efficient than jumping straight to E-3 + Maxx.
