Disclosure: some links on this page may be affiliate links; we may earn a commission at no extra cost to you. Read the disclosure.
By the toolist.ai editors. Read from Devin's pricing page, not hands-on tested by toolist.ai. Updated October 16, 2026.
Devin pricing
Devin's cheapest paid plan is Pro at $20 a month billed monthly. The free plan: Free gives unlimited Tab completions and inline edits in Devin Desktop plus a light agent quota with limited models; Cognition publishes no figure for the quota, the plan has no Devin Cloud, Devin API or integrations, and when the quota runs out you wait for the daily or weekly reset.
Free plan: Free gives unlimited Tab completions and inline edits in Devin Desktop plus a light agent quota with limited models; Cognition publishes no figure for the quota, the plan has no Devin Cloud, Devin API or integrations, and when the quota runs out you wait for the daily or weekly reset.
| Plan | Billed monthly | Billed yearly, per month | Includes |
|---|---|---|---|
| Free | free | not listed | Devin Desktop with unlimited Tab completions and inline edits, a light agent quota that resets daily and weekly, and limited models without the premium ones (Cognition's models doc names SWE-1.6 as the only model on Free); no Devin Cloud, Devin API, Ask Devin or integrations |
| Pro | $20 | not listed | One user: a daily and weekly quota shared by Devin Cloud, Devin Desktop and the CLI; frontier models from OpenAI, Anthropic, Google and SpaceXAI plus open-source models; Devin Cloud with up to 10 concurrent sessions; Slack, Microsoft Teams, Linear, Jira, GitHub, GitLab and Bitbucket; extra usage at API prices |
| Max | $200 | not listed | One user: everything in Pro with a significantly larger weekly quota and no daily cap (Cognition gives no figure), shared by Devin Cloud, Desktop and the CLI; up to 10 concurrent sessions |
| Teams | $80 | not listed | Up to 200 users with an $80 monthly minimum: full seats cost $40 a month each and include a Pro-sized quota and Devin Desktop, free flex seats use the team's shared prepaid credits only; unlimited concurrent sessions, sharing and collaboration, central billing, an admin dashboard with analytics, priority support and a Slack Connect support channel; no SSO |
| Enterprise | not listed | not listed | Contact sales: billed in Agent Compute Units (ACUs) at the order-form rate, with organization and per-user ACU limits; SAML/OIDC SSO, deployment in your own cloud (VPC), custom git providers, management of several organizations, a dedicated account team, and no training on your data without written consent |
| Extra usage (on-demand credits) | not listed | not listed | Prepaid credits on paid plans, used after the quota at API list prices for the model and never expiring; per million tokens input/output in Devin Desktop and the CLI, for example Claude Opus 5.5 $4/$20, Claude Sonnet 5.5 $2/$10, GPT-6.1 Sol $2/$10 and SWE-1.6 $0.50/$2.50, with the fast variants of those Claude and GPT models at twice the price; optional auto-reload, shared across a Teams account |
| Cost per finished minute | Not measured by us, and the vendor does not publish an allowance we can divide by. |
|---|
Prices last checked: 2026-10-16 on Devin's pricing page.
Price history
No price change since we first recorded these prices on October 14, 2026. We re-check them every week and list each change here.
To price your own month on Devin and the tools like it, use the AI cost calculator.
Before you pay
- Cognition publishes no quota figures: each paid plan's allowance refreshes daily and weekly (Max weekly only, with no daily cap), and what a message uses depends on the model, the size of the task and the reasoning. Cognition says most users never reach their limits; when you do, Free waits for the reset and paid plans buy extra usage, billed at API list prices for the model, with fast and priority variants costing more.
- Devin Cloud sessions are metered by the work Devin does (its actions plus virtual-machine time), not per task, so a long or vague task uses more; Windows sessions use about 9% more than Linux, and a session goes to sleep after 30 minutes idle. Enterprise is billed in Agent Compute Units (ACUs) at a contract rate with no public price.
- Teams pricing reads two ways: the pricing page says '$80/month for team plan + $40/mo per full dev seat', while Cognition's billing docs say the $80 is a minimum that full seats count toward (two full seats cost $80 in all, three $120, and with fewer than two the rest is charged as prepaid credits). Free flex seats draw only on the team's shared credits and do not include Devin Desktop, and new Teams plans no longer include SSO, which is Enterprise-only.
- Former Windsurf users: the rename on June 2, 2026 left plans and prices unchanged, but Windsurf had already replaced prompt credits with daily and weekly quotas in March 2026. Pro subscribers who paid $15 a month keep that price indefinitely and Teams subscribers keep $30 a Developer seat, while new subscribers pay $20 for Pro and $40 a full Teams seat; unused add-on credits became extra-usage balance at $10 per 250 credits, refundable through support. Legacy Windsurf Enterprise customers keep credit billing ($120 per 1,000 pooled credits) and the windsurf.com login.
- The old Cascade agent was replaced by Devin Local (Cognition's launch post kept Cascade for migration through July 1, 2026, and its release notes show enterprise admins can still switch legacy Cascade on), the Windsurf JetBrains plugin is in maintenance mode with Cognition pointing JetBrains users to Devin over the Agent Client Protocol, and the app is now named Devin on every system, so company device-management allowlists that approve only 'Windsurf' may block it.
- SWE-2, Cognition's own model, was free in Devin Desktop and the CLI on paid plans only under a launch offer running to mid-October 2026 (October 16 on the pricing page, October 15 in the models doc); after it, SWE-2 is billed against your quota at its list price of $3 input and $15 output per million tokens ($0.30 cached input).
- By default Cognition may use your code and prompts to train its models: paid plans can opt out in Data Controls (on Teams only an admin can), which also turns on zero data retention with model providers, the Free plan cannot opt out, and Enterprise data is not used without written consent.
- Payments are non-refundable except where Cognition's terms or the law say otherwise, though its docs say you can ask for a refund before using a new subscription; plans renew automatically, a cancelled plan runs to the end of the period, Pro and Max are single-user plans, and the pricing page lists monthly prices only. Free trials go only to customers Cognition's systems pick, and generally not to anyone who has used Devin, Windsurf or Codeium before.
See the full Devin review and how we research.
Sources
- Devin plans and pricing (checked 2026-10-14)
- Devin product page (checked 2026-10-14)
- Devin Desktop product page (checked 2026-10-14)
- Devin blog: Windsurf is now Devin Desktop (checked 2026-10-14)
- Cognition blog: Introducing SWE-2 (checked 2026-10-14)
- Devin docs: Billing (checked 2026-10-14)
- Devin docs: Self-serve plans (checked 2026-10-14)
- Devin docs: Usage (checked 2026-10-14)
- Devin docs: Enterprise billing (checked 2026-10-14)
- Devin docs: Use your ChatGPT plan in Devin (checked 2026-10-14)
- Devin docs: Devin Desktop FAQ (Windsurf to Devin Desktop) (checked 2026-10-14)
- Devin docs: Devin Desktop plans and usage (checked 2026-10-14)
- Devin docs: Quota-based usage (migration for Windsurf subscribers) (checked 2026-10-14)
- Devin docs: AI models and token prices in Devin Desktop (checked 2026-10-14)
- Devin docs: Devin Review (checked 2026-10-14)
- Devin docs: Security and data use (checked 2026-10-14)
- Devin docs: 2026 release notes (checked 2026-10-14)
- Devin docs: 2025 release notes (Devin 2.0 on April 3, 2025) (checked 2026-10-14)
- Cognition Platform Terms of Service (checked 2026-10-14)
- Zapier: Devin Desktop (formerly Windsurf) vs. Cursor (checked 2026-10-14)
- Answer.AI: Thoughts on a month with Devin (checked 2026-10-14)
- Builder.io on DEV Community: Devin review, is it a better AI coding agent than Cursor? (checked 2026-10-14)
- Trustpilot: Windsurf (windsurf.com) (checked 2026-10-14)