Obvious/Help Center

Agent Modes (Fast, Auto, Deep)

Published March 4, 2026 · Last updated March 5, 2026 · 4 min read

Every conversation in Obvious runs through an agent mode — a combination of AI model, reasoning depth, and speed that shapes how your agent works. You choose the mode before (or during) a conversation, and Obvious handles the rest: routing your request to the right model with the right settings.

The mode selector sits at the bottom-left of the chat input. Click it to see your available modes, or use Ctrl+1 through Ctrl+9 to switch instantly by position. Ctrl+Shift+[ and Ctrl+Shift+] cycle through modes one at a time.

Auto

Auto is the default mode, and for most work, it's the right one. It runs on Claude Sonnet 4.6 with adaptive thinking — meaning the agent decides how hard to reason based on the complexity of your request. A simple rename gets a fast response. A multi-step data transformation gets deeper analysis.

Sonnet 4.6 delivers faster responses while handling the full range of everyday tasks — creating documents, editing sheets, running queries, building workflows. If you're not sure which mode to use, start here. You can always switch mid-project if you need more depth or less latency.

Fast

Fast mode runs on Claude Haiku 4.5 — a smaller, faster model that responds in a fraction of the time. It's ideal for rapid-fire conversations: quick edits, simple lookups, formatting tasks, or situations where you need to iterate fast and don't need deep reasoning.

The trade-off is straightforward. Fast mode won't plan multi-step workflows as carefully, and it may miss nuances in complex requests. But for the volume of everyday tasks that don't require heavy lifting, it's noticeably quicker.

Deep Work

Deep Work mode switches to GPT-5.2 with high reasoning effort. This is the mode for problems that benefit from sustained, methodical thinking: complex data analysis, multi-source research, detailed writing projects, or anything where getting the first answer right matters more than getting it quickly.

Deep Work takes longer per response. That's the point — the model spends more time reasoning before it responds, which produces more thorough and more accurate results on challenging tasks.

Deep Work (Opus)

Deep Work (Opus) is the maximum-capability mode, running Claude Opus 4.6 with the highest reasoning effort Obvious supports. Use it for the most demanding tasks: long-form synthesis, intricate multi-step workflows, or work where even small mistakes are costly. This mode uses significantly more compute, so it's reserved for moments that justify it.

Analyst

Analyst mode is purpose-built for quantitative work. It runs on GPT-5.1 with high reasoning effort and is tuned specifically for data analysis tasks: exploring datasets, building queries, identifying trends, and producing charts. If your work revolves around numbers and patterns, Analyst mode is worth trying before you reach for Deep Work.

Choosing the Right Mode

Most decisions come down to a simple question: how complex is this task?

For quick edits, lookups, and simple formatting, use Fast. For everyday work that mixes creating, analyzing, and editing, use Auto. For complex analysis, research, or writing that benefits from deeper reasoning, use Deep Work or Analyst. For the highest-stakes tasks where precision is everything, use Deep Work (Opus).

You can switch modes at any point in a conversation. The mode applies to the next message you send — it doesn't retroactively change previous responses. Experiment freely. The mode selector remembers your last choice per project, so you won't need to reset it every session.

Next Steps

Was this helpful?