Devin Pricing 2026: Plans, ACU Costs, and Whether It's Worth It
Devin, Cognition’s autonomous AI software engineer, uses a usage-based pricing model built around the ACU — Agent Compute Unit. That makes it cheap to try and hard to predict, so the real question is not the sticker price but how many ACUs your tasks actually burn. Here is the full breakdown of the three plans and what they cost in practice.
The three Devin plans
Cognition sells Devin on three plans that span a wide range:
| Plan | Price | Included usage | Best for |
|---|---|---|---|
| Core | ~$20/month | Pay-as-you-go, ~$2.25 per extra ACU | Solo devs testing Devin on light tasks |
| Team | ~$500/month | 250 ACUs included (~$2.00 each) | Teams running Devin regularly with unlimited concurrency |
| Enterprise | Custom | Custom, VPC deployment | Orgs needing SSO, isolation, and their own infrastructure |
The Core plan replaced Devin’s old $500-only entry point and exists to let individuals try the product without a big commitment: you pay a small base and then buy compute as you go. Team bundles a block of usage at a lower per-ACU rate plus unlimited concurrency, which is the meaningful upgrade if you want to run many Devins at once. Enterprise adds VPC deployment, SAML/OIDC SSO, teamspace isolation, and a dedicated account team. Always confirm the current numbers on the official pricing page and billing docs, since Cognition adjusts them periodically.
What an ACU actually is
An ACU is Cognition’s normalized unit of the resources Devin consumes while actively working — virtual machine time, model inference, and network bandwidth rolled into one number. The rule of thumb Cognition gives:
1 ACU ≈ 15 minutes of active, autonomous Devin work.
“Active” is the key word. You are billed for the time Devin spends planning, writing, running tests, and browsing, not for wall-clock time it sits idle waiting on you. A quick, well-scoped bug fix might cost a fraction of an ACU; a sprawling migration where Devin thrashes on a flaky test suite can eat several.
What Devin actually costs per task
Because billing is consumption-based, your real cost depends entirely on task complexity. Rough ranges people report:
- Small, tightly scoped tasks (a clear bug, a small endpoint) — often under 1 ACU, so roughly $2 or less
- Medium features behind an existing pattern — a few ACUs, roughly $5–$15
- Large or ambiguous work — can run 5–10+ ACUs, and can fail without producing a mergeable PR
That last case is the one to watch. When a task is under-specified, Devin can burn ACUs iterating on the wrong thing and still hand back something you cannot merge. On Core’s pay-as-you-go model those charges accumulate quietly, which is the same dynamic that has drawn scrutiny at Replit Agent, where effort-based pricing made costs hard to predict up front. The fix is the same on Devin: scope tightly, watch the ACU meter, and set spend limits.
Is Devin worth it
Devin is worth it when your work matches its strengths — a backlog of well-defined, parallelizable tickets that a real engineer would otherwise context-switch to handle. In that mode, ~$2 per fifteen minutes of autonomous engineering that opens a reviewable PR is cheap leverage. Run several Devins on Team’s unlimited concurrency and you can clear a backlog while your team ships the hard stuff.
It is a poor fit when your work is ambiguous, exploratory, or architecture-heavy. There, Devin burns ACUs producing confident, wrong output and you pay for the privilege of reviewing it. The break-even is less about the price and more about how disciplined your task scoping is.
If cost predictability matters more than autonomy, an editor-based tool with flat seat pricing may fit better — see Devin vs Cursor and our Devin alternatives roundup, which compares the ACU model against per-seat and per-request pricing across the field. For a sense of where autonomous agents sit alongside app builders, our best AI tools for founders guide is a useful map.
The cost Devin doesn’t cover
Every ACU you spend goes toward shipping code. None of it goes toward running the business around that code — the customer emails, the scheduling, the follow-ups that pile up whether or not the feature ships. That work is a different job, and an AI executive assistant like Carly handles it: it works over email with no app to install, gives each agent its own email address, custom instructions, and memory, and connects to 200+ integrations across 40+ categories. Pricing starts at $35/month. Devin bills for building; Carly handles the operations that building creates.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
