GenGrades

Devin review: pricing, SWE-2 Flash and what public tests found

Published Updated

Devin logodevin.aiSubscriptionWebAPIDesktopCLIPlugin
The Devin homepage: "Meet Devin, your team's autonomous software engineer", under a banner offering enterprises more than $100 million in Devin Security credits.
The Devin homepage: "Meet Devin, your team's autonomous software engineer", under a banner offering enterprises more than $100 million in Devin Security credits.Screenshot of devin.aitaken

What Devin is

In one or two sentences, before anything else.

Devin is Cognition's autonomous coding agent: it works on its own cloud computer and opens pull requests, and ships with Devin Desktop (formerly Windsurf) and a CLI. In November 2026 Cognition added the cheaper SWE-2 Flash model and Devin Review 2.0. Plans cost US$0 to US$200 a month; the best-known outside test, Answer.AI's, saw 3 successes in 20 tasks.

Main features

What it can do, one sentence each, and whether we actually tested it.

  • Devin Cloud: an agent with its own computer

    Takes a task from the web app, Slack, Microsoft Teams, Linear, Jira or the API, works on its own Linux, Windows or macOS virtual machine with a shell, browser and editor, clicks through what it built and opens a pull request with a screen recording.

    Per Devin's homepage and computer-use docs. The pricing page allows up to 10 cloud sessions at once on Pro and Max, unlimited on Teams and Enterprise; the Free plan has no Devin Cloud.

  • Devin Desktop, formerly Windsurf

    An editor built on Windsurf whose Agent Command Center lists local and cloud agents on one board. It runs Devin Local and other agents that speak the Agent Client Protocol, such as Codex, Claude Agent and OpenCode, and keeps unlimited Tab completions on every plan.

    Renamed on June 2, 2026, per Cognition's announcement; for macOS, Windows and Linux per the download page.

  • Devin CLI

    A local agent in the terminal on macOS, Linux, WSL and Windows that can hand a task to a cloud Devin and bring the finished pull request back to your machine.

    Per Devin's CLI docs. It also runs inside JetBrains IDEs, Zed and Xcode through the Agent Client Protocol.

  • Devin Review

    Reviews GitHub and GitLab pull requests from any author: it orders the diff, flags likely bugs by severity, can re-review only what changed on each push and hands findings to Devin to fix. Public GitHub PRs can be reviewed free at devinreview.com.

    Per Cognition's November 5, 2026 post: a typical review costs US$0.25 to US$0.75 and median time fell to 94 seconds. These are the maker's figures.

  • SWE-2 and SWE-2 Flash, Cognition's own models

    SWE-2 is post-trained from Moonshot's Kimi K3; SWE-2 Flash is built on Thinking Machines' Inkling-Small. Both come in Medium, High and Max effort. SWE-2 Flash runs only in Devin Desktop and the CLI, not in Devin Cloud.

    Per Cognition's posts of September 10 and November 4, 2026 and the models page. Their benchmark scores are Cognition's own, several on its own FrontierCode benchmark.

  • Other labs' models, Fusion and Adaptive

    The self-serve price list has more than 250 model and effort options, from Claude Fable 5.1 and GPT-6.1 Sol to Moonshot's Kimi K3.1 Pro, Zhipu's GLM-5.3 and DeepSeek V4. Fusion pairs a frontier lead model with a cheaper sidekick; Adaptive picks the model for you.

    Price list dated November 5, 2026. Cognition claims Fusion matches frontier scores at up to 60% lower cost on FrontierCode, its own benchmark.

  • Automations, Oncall and Security

    Starts sessions on its own from Slack, GitHub, Linear, PagerDuty, schedules or webhooks to triage bugs, fix failing CI, investigate alerts and patch vulnerabilities reported by scanners such as SonarQube, Snyk and Wiz.

    Devin Oncall became generally available on October 15 and Devin Security 2.0 launched on October 26, 2026, per Cognition's blog.

  • Enterprise deployment options

    Enterprise Cloud runs every session on its own isolated machine; Customer Dedicated Deployment puts Devin's workspace in a single-tenant cloud network reached over AWS PrivateLink; Outposts run sessions on your own machines. Devin's "brain", the service that powers its reasoning, always runs in Cognition's cloud.

    Per Devin's enterprise deployment docs.

Who it is for

The people it helps and the problems it solves, and where it ranks.

Built for

  • Engineering teams with a backlog of small, clearly specified tickets: bugs, tests, dependency upgrades
  • Enterprises running one migration or vulnerability fix across dozens or hundreds of repositories
  • Teams that want to start work from Slack, Microsoft Teams, Linear or Jira and get pull requests back
  • Windsurf users, who now get Devin Desktop, the CLI and cloud agents on one plan
  • Teams drowning in pull requests, including ones written by other agents, who want an automated first review

Problems it solves

  • Getting well-defined tasks done without a developer at the keyboard
  • Reviewing the growing volume of pull requests before they merge
  • Running the same change across many repositories in parallel
  • Capping AI coding spend per user, team or pull request

What it is good at

Public tests other people ran: what they gave it and how it did.

  1. Case 1

    Answer.AI's month with Devin: Hamel Husain, Isaac Flath and Johno Whitaker gave it 20 real tasks and published the results on January 8, 2025

    Input

    Pull data from a Notion database into Google Sheets; scrape an author's 25 latest papers from Google Scholar with Playwright; deploy several apps to one Railway deployment; migrate a Python project to nbdev.

    What happened

    3 tasks succeeded, including the Notion-to-Sheets script, 14 failed and 3 were inconclusive. Devin spent more than a day on the Railway task, which Railway does not support, and the testers could not predict which tasks would work. This was the 2024 version of Devin, before Devin 2.2's computer use and the SWE-2 models.

  2. Case 2

    Security researcher Johann Rehberger (Embrace The Red) paid US$500 for a month of Devin in April 2025 to test prompt injection, and published in August 2025

    Input

    Ask Devin to investigate the latest open GitHub issue, where the issue text points to a 'support tool' on an attacker's website.

    What happened

    Devin followed the planted instructions, downloaded a binary, made it executable and ran it, giving the researcher remote control of its machine. Follow-up posts showed secrets leaking through the shell and browser and a tool that can expose a local port to the internet. He reported it on April 6, 2025 and had no fix timeline 120 days later; Devin now offers security profiles that limit network, MCP and git access.

  3. Case 3

    Tenki, which sells a competing AI code reviewer, ran seven tools on the same pull requests in April 2026

    Input

    50 pull requests from cal.com, Sentry, Grafana, Keycloak and Discourse that reintroduce 122 real, later-fixed bugs, each finding graded by three LLM judges.

    What happened

    Devin caught 44 of 122 bugs (36.1% recall) at 47.3% precision, an F1 of 40.9 against Tenki's own 41.7; Greptile caught 44, CodeRabbit 35 and GitHub Copilot 30. Tenki lists Devin as a coding agent shown for reference, and the run predates Devin Review 2.0.

  4. Case 4

    Per Cognition's Devin Review 2.0 post (November 5, 2026): reviewing every pull request automatically

    Input

    An auto-review rule that runs on every push once a pull request is ready, limited to files under src/billing/**, with a per-PR spend limit.

    What happened

    Cognition says a typical review costs US$0.25 to US$0.75, about 80% of pull requests cost under a dollar including every push, and a fixed bug costs about US$2.25 on average across its customers. These are the maker's numbers; no outside measurement of Review 2.0 was found.

  5. Case 5

    The Internet of Bugs YouTube channel went through Devin's March 2024 launch demo of an Upwork job frame by frame (April 9, 2024)

    Input

    Cognition's March 2024 demo in which Devin takes an Upwork job to write and debug code that runs a computer-vision model from a GitHub repository.

    What happened

    The video argues that Devin did not deliver what the client asked, invented errors and then fixed them, and took at least six hours on work the author redid in about 36 minutes. It concerns the 2024 demo, not today's product, but it is why many developers still distrust Devin's claims.

Where it falls short

What it does badly, with the evidence, and what we could not settle.

Weak

Weak

  • Independent evidence is old or thin.

    Evidence: The best-known outside test, Answer.AI's in January 2025, counted 3 successes, 14 failures and 3 inconclusive results in 20 tasks. No third-party SWE-bench-style results for Devin itself were found; Cognition's 2026 scores for SWE-2 and SWE-2 Flash are its own, several on its own FrontierCode benchmark.

  • A month's cost is hard to predict.

    Evidence: Cognition does not state Pro or Max quotas in tokens or tasks; use depends on the model, task size, session length and back-and-forth, and anything past the quota is billed at API token prices. SWE-2's free period in Desktop and the CLI ends November 30, 2026, after which the docs list US$3 per million input tokens and US$15 per million output. A Hacker News commenter who builds a competing product said Devin Automations 'can get very expensive to run'.

  • Your code may be used to train Cognition's models.

    Evidence: Section 3.3.1 of the terms updated June 30, 2026 allows training on customer data by default. Paid plans can opt out (on Teams only an admin can); the Free plan cannot. Enterprise data is not used without written consent.

  • Refunds are narrow and there is no trial of paid plans.

    Evidence: The terms say payments are non-refundable except where the terms or the law say otherwise; Devin's docs offer a refund only if you ask before using the subscription, and say free trials of paid plans are not offered.

  • An unattended agent can be hijacked by what it reads.

    Evidence: In 2025 Embrace The Red showed instructions planted in a GitHub issue making Devin download and run malware and leak secrets; at the time Devin had unrestricted internet access by default. Cognition's own security docs warn that Devin can introduce bugs or suggest insecure code.

  • Surfaces and models do not behave alike.

    Evidence: In the September 2026 Hacker News thread on SWE-2, users said the web and desktop clients are different harnesses with different quality and called the CLI buggy. SWE-2 Flash is not available in Devin Cloud, and SWE-2's free period covers only Desktop and the CLI.

Uncertain

Uncertain

  • How large the Pro and Max quotas are.

    Why unsure: The pricing page says only 'increased quotas' and 'significantly higher quotas' and that most users will never hit their limits, which is Cognition's own claim.

  • When SWE-2 Flash stops being free.

    Why unsure: Cognition's launch post says 'Free through 11/30', while the model price list dated November 5, 2026 says free until December 4, 2026 at 18:00 UTC.

  • Whether Devin Review 2.0's savings hold outside Cognition's customer data.

    Why unsure: The 82% cost cut, 1.32 bugs per PR and US$2.25 per fixed bug are Cognition's averages; Tenki's comparison was run before version 2.0.

  • What the Free plan includes.

    Why unsure: The billing docs list Devin Review and DeepWiki on Free, but the pricing page's comparison table marks DeepWiki, Ask Devin and the API as not included.

Price, limits and formats

The numbers, each with its source and the date we checked it.

SpecValueNote and source
MakerCognition AI, Inc., San Francisco

Devin was introduced on March 12, 2024; Cognition bought the Windsurf editor in July 2025.

Source: cognition.comchecked

SurfacesDevin Cloud (web app, Slack, Teams, API), Devin Desktop for macOS, Windows and Linux, Devin CLI, JetBrains plugin, Devin Review; mobile app on a waitlist only

Source: devin.aichecked

Free planUS$0: light quota, limited models, unlimited Tab and inline edits; no Devin Cloud, API or Ask Devin; SWE-2 at slow speed

Source: devin.aichecked

Pro planUS$20 a month, one user, daily and weekly quota, Devin Cloud with up to 10 sessions at once

Source: devin.aichecked

Max planUS$200 a month, one user, larger weekly quota with no daily cap

Source: docs.devin.aichecked

Teams planAt least US$80 a month; US$40 a month per full seat with Pro-level quota; free flex seats on shared credits; up to 200 users

Source: docs.devin.aichecked

EnterpriseCustom price, billed in Agent Compute Units (ACUs); SAML/OIDC SSO, dedicated deployment

Source: docs.devin.aichecked

Use past the quotaPrepaid on-demand credits charged at API token prices; credits never expire and can auto-reload

Source: docs.devin.aichecked

SWE 2 priceFree on self-serve plans in Desktop and CLI through November 30, 2026 (not in cloud sessions); then US$3 per million input and US$15 per million output tokens

Source: docs.devin.aichecked

SWE 2 Flash priceFree until December 4, 2026, 18:00 UTC; then US$0.10 per million input and US$0.50 per million output tokens; Desktop and CLI only

Source: docs.devin.aichecked

Devin Review costTypical review US$0.25 to US$0.75 per pull request, per Cognition; public GitHub PRs free at devinreview.com

Source: devin.aichecked

MeteringSessions sleep after 5 minutes idle and use nothing while asleep; Windows sessions use about 9% more than Linux

Source: docs.devin.aichecked

RefundsPayments non-refundable except as the terms or law provide; refund on request only before the subscription is used

The refund-before-use rule and 'no free trials' are in Devin's Plans and Usage docs.

Source: cognition.comchecked

Training on your dataAllowed by default; paid plans can opt out (Teams: admins only); Enterprise data not used without written consent

Source: cognition.comchecked

CertificationsSOC 2 Type II since September 2024; ISO 27001; FedRAMP High for Windsurf extensions since March 2025

SOC 2 date from Devin's security docs; ISO 27001 and FedRAMP badges on the homepage; FedRAMP High per Cognition's March 10, 2025 post.

Source: docs.devin.aichecked

Where code runsEach session on its own isolated virtual machine in Cognition's cloud on AWS; single-tenant network or your own machines (Outposts) for Enterprise

Source: docs.devin.aichecked

ModelsMore than 250 model and effort options on self-serve plans, including Claude, GPT, Gemini, Kimi K3.1 Pro, GLM, DeepSeek and the SWE family

Source: docs.devin.aichecked

FundingMore than US$2 billion Series E at a US$48 billion valuation, September 8, 2026, led by Andreessen Horowitz and Accel

Confirmed by Dealroom and Tech Funding News; the previous round, in May 2026, valued Cognition at US$26 billion.

Source: cognition.comchecked

RevenueUS$1 billion annualized run rate on September 25, 2026, per Cognition

Source: cognition.comchecked

Incidents25 incidents on Devin's status page from May 13 to November 4, 2026, including a web-app outage on September 24

Source: devinstatus.comchecked

Common questions

Answer first, then why, an example, the limits and what to do next.

What is Devin AI?

Devin is an AI coding agent from Cognition that takes a task, works on its own cloud computer and returns a pull request. The same subscription now includes Devin Desktop (the editor formerly called Windsurf), the Devin CLI for your terminal and Devin Review for checking pull requests.

Why

Cognition bought Windsurf in July 2025 and renamed it Devin Desktop on June 2, 2026, so one account covers local and cloud agents.

Is Devin free?

Devin has a free plan, but it covers only Devin Desktop and the CLI with a light quota and limited models; Devin Cloud, the API and Ask Devin start with Pro at US$20 a month. Free users can run SWE-2 Flash Medium, which costs nothing until December 4, 2026 per Devin's model price list.

Limits

Free plans cannot opt out of Cognition using their data for training, and paid plans have no free trial.

How much does Devin cost?

Devin costs US$20 a month for Pro and US$200 for Max, both for one person, and at least US$80 a month for Teams, where each full seat is US$40. Every paid plan has a daily or weekly quota; past it you pay with prepaid credits at API token prices, and Enterprise is priced on request in ACUs.

Why

Cognition replaced the old US$500 Team plan and the pay-per-ACU Core plan with these plans on April 14, 2026.

Limits

Cognition does not state how much work a quota covers, so heavy use is hard to price in advance.

Does Devin still charge in ACUs?

Only Enterprise customers are billed in Agent Compute Units (ACUs), at the rate in their order form. Since April 2026, Free, Pro, Max and Teams use a quota that refreshes daily and weekly, and extra use is billed in dollars at each model's token price.

What are SWE-2 and SWE-2 Flash, and are they free?

SWE-2 and SWE-2 Flash are Cognition's own coding models, and both are free on self-serve plans for a limited time. SWE-2, released September 10, 2026 and post-trained from Kimi K3, is free in Devin Desktop and the CLI through November 30, 2026; SWE-2 Flash, released November 4 and built on Thinking Machines' Inkling-Small, is free until December 4.

Why

Cognition says SWE-2 Flash costs one-fifth as much as SWE-2 and runs three times as fast; its benchmark scores are Cognition's own.

Limits

Neither free period covers Devin Cloud sessions, and SWE-2 Flash is not offered in Devin Cloud at all.

Is Devin any good? What do independent tests show?

Independent tests of Devin are few and mostly old. Answer.AI's month-long test, published in January 2025, found 3 successes in 20 tasks; in a 2026 code-review benchmark run by Tenki, a competitor, Devin caught 44 of 122 bugs, the second-best F1 score of seven tools.

Why

Cognition reports that 67% of Devin's pull requests were merged in late 2025, up from 34% a year earlier, but that figure is its own.

Next step

Its docs say Devin works best on tasks you could finish in three hours or less with clear success criteria; start there.

Does Devin train on my code?

Devin's terms let Cognition use your code and prompts to train models unless you are on a paid plan and opt out. After an opt-out, Cognition says your data is not used for training and zero data retention is switched on with its model providers; on Teams only an admin can opt out, and Enterprise data is never used without written consent.

Is Devin safe to connect to my repositories?

Devin is built with standard enterprise controls, but an autonomous agent with repository and internet access needs guardrails. Cognition reports SOC 2 Type II since September 2024 and ISO 27001, runs each session on its own virtual machine, and lets you choose which GitHub repositories it can see.

Why

In 2025 a security researcher showed that instructions hidden in a GitHub issue could make Devin run malware and leak secrets.

Next step

Use a security profile with a network allowlist, keep secrets in Devin's secrets store and require branch protection before merges.

Devin vs Claude Code, Codex or Cursor: which should I use?

Choose Devin to hand whole tickets to a cloud agent that tests on its own machine and reports back in Slack or Linear; choose Claude Code or Codex if you mostly want to drive an agent yourself inside a plan you already pay for. Claude Code comes with Claude Pro at US$20 a month, Codex with ChatGPT Plus at US$20, and Cursor's cloud agents start with its US$30 Premium plan.

Limits

Quotas differ by maker and are not comparable; try the same task in two of them before committing a team.

What happened to Windsurf?

Windsurf is now Devin Desktop. Cognition, which bought Windsurf in July 2025 after Google hired its chief executive, renamed the editor on June 2, 2026 as an over-the-air update that kept users' plans, settings and extensions.

Why

Windsurf had already moved from prompt credits to daily and weekly quotas in March 2026; existing US$15 Pro subscribers keep that price.

Can I get a refund from Devin?

Devin refunds a subscription only if you ask before using it. Cognition's terms say payments are otherwise non-refundable unless the terms or the law require it, and its docs say free trials of paid plans are not offered.

Limits

Unused on-demand credits never expire, so they stay usable after a plan change.

How much does Devin Review cost?

Cognition says a typical Devin Review costs US$0.25 to US$0.75 per pull request, drawn from your plan's quota or credits, and reviews of public GitHub pull requests at devinreview.com are free. A per-PR spend limit stops a runaway review from running up the bill.

Why

Cognition says it cut the cost of a review by 82% between August and November 2026 by mixing cheaper models.

Limits

The Ultra effort level runs a more thorough review at a higher price, and on Teams reviews draw from the shared credit pool.

More screenshots

  • The pricing page: Free $0, Pro $20 a month and Max $200 a month for individuals; Teams starts at $80 a month and Enterprise is priced on request.
    The pricing page: Free $0, Pro $20 a month and Max $200 a month for individuals; Teams starts at $80 a month and Enterprise is priced on request.Screenshot of devin.aitaken
  • The plan comparison: Free has no Devin Cloud, DeepWiki, API or Ask Devin and only slow free SWE-2; Pro and Max run up to 10 cloud sessions at once, Teams and Enterprise unlimited.
    The plan comparison: Free has no Devin Cloud, DeepWiki, API or Ask Devin and only slow free SWE-2; Pro and Max run up to 10 cloud sessions at once, Teams and Enterprise unlimited.Screenshot of devin.aitaken
  • Devin's self-serve price list for use beyond the quota: SWE-2 Flash free until December 4, 2026 at 18:00 UTC, then $0.10 per million input tokens and $0.50 per million output.
    Devin's self-serve price list for use beyond the quota: SWE-2 Flash free until December 4, 2026 at 18:00 UTC, then $0.10 per million input tokens and $0.50 per million output.Screenshot of docs.devin.aitaken

Pros and cons

The short version of everything above.

Pros

  • A cloud agent on its own machine that tests its work and opens pull requests from Slack, Linear or Jira
  • One plan covers the cloud agent, a desktop editor, a CLI and code review
  • Cognition's SWE-2 models are free on self-serve plans for now, beside 250+ other model options
  • Spending controls: per-PR review limits, auto-reload caps and enterprise budgets
  • Enterprise options: SOC 2 Type II, ISO 27001, single-tenant cloud or sessions on your own machines

Cons

  • Little recent independent testing; the best-known outside trial found 3 successes in 20 tasks
  • Quotas are not published in tokens or tasks, so a month's cost is hard to predict
  • Code may train Cognition's models by default, and Free users cannot opt out
  • No free trial of paid plans and refunds only before a subscription is used
  • Prompt-injection research showed an unattended Devin could be made to run malware

What the public tests show

This review is built from what could be checked on November 6, 2026: Devin's and Cognition's own pages (pricing, documentation, the model price list, changelogs, blog posts, terms, privacy policy and status page); a month-long test by Answer.AI; an analysis of the launch demo by the Internet of Bugs YouTube channel; prompt-injection research by Embrace The Red; a code-review benchmark run by Tenki, which sells a competing reviewer; Hacker News threads; reporting by TechCrunch, Built In, Dealroom, Tech Funding News and Channel Insider; and rivals' pricing pages. Cognition's claims about itself are labelled as its own.

What Devin is

Devin is the AI coding agent of Cognition AI, Inc., a San Francisco company that introduced Devin on March 12, 2024 as "the first AI software engineer". In 2026 Devin is less one product than a family under one account:

  • Devin Cloud, the original agent. You give it a task in the web app, Slack, Microsoft Teams, Linear, Jira or through the API; it works on its own Linux, Windows or macOS virtual machine, clicks through what it built and opens a pull request with a screen recording.
  • Devin Desktop, the editor Cognition got when it bought Windsurf in July 2025, renamed on June 2, 2026. Its Agent Command Center shows local and cloud agents on one board and also runs outside agents such as Codex, Claude Agent and OpenCode.
  • Devin CLI, a local agent for the terminal that can hand a task to a cloud Devin and pull the finished branch back.
  • Devin Review, a pull-request reviewer for code written by people or by agents.

Cognition is heavily funded. It raised more than US$2 billion at a US$48 billion valuation on September 8, 2026, led by Andreessen Horowitz and Accel, as Dealroom and Tech Funding News also reported, and says it passed US$1 billion in annualized revenue on September 25. Tech Funding News and Channel Insider, citing The Information, reported that its cash burn could reach about US$800 million this year.

What changed in October and November 2026

  • SWE-2 Flash (November 4). Cognition's new small coding model is built on Inkling-Small from Thinking Machines, which Cognition calls a fully American base, and comes in Medium, High and Max effort. Cognition says it costs one-fifth as much as SWE-2 and runs three times as fast, and scores 24.0% on version 2.0 of FrontierCode, Cognition's own benchmark, against 23.8% for SWE-2. It runs only in Devin Desktop and the CLI. Cognition is not yet releasing the weights publicly, citing the risk of misuse in cyber attacks.
  • Devin Review 2.0 (November 5). Cognition says the cost of reviewing a pull request fell 82% since August, to US$0.25 to US$0.75 for a typical review, while bugs found per PR rose from 1.22 to 1.32 and median time fell from 139 to 94 seconds. New are incremental reviews of each push, an Ultra effort level and custom reviewers in beta.
  • Kimi K3.1 Pro. Moonshot's Kimi K3.1 Pro appears in Devin's model price list, dated November 5, 2026, in four effort levels at US$2 per million input tokens and US$10 per million output. No announcement post was found.
  • Earlier this autumn: SWE-2 on September 10; sign-in with ChatGPT on September 29, which lets a ChatGPT Plus or Pro plan pay for GPT models inside Devin; Devin Oncall on October 15; Devin Security 2.0 on October 26, with more than US$100 million in credits offered to enterprises; and hierarchical spending budgets on November 2.

SWE-2 is post-trained from Moonshot's Kimi K3, a 2.8-trillion-parameter open model. Cognition says it scores 50.0% on FrontierCode 1.1 Main, within a point of Claude Fable 5.1 at 64% lower cost. Every one of these numbers comes from Cognition, and FrontierCode is Cognition's own test.

How Devin is priced

Cognition rebuilt its self-serve pricing on April 14, 2026. The US$500-a-month Team plan and the pay-per-ACU Core plan went away, and Ask Devin, DeepWiki and Devin Review stopped being free. This is the lineup on the pricing page:

PlanPriceWho and whatDevin Cloud
FreeUS$0One user; light quota, limited models, unlimited Tab completionsNo
ProUS$20 a monthOne user; daily and weekly quota; frontier models; SWE-2 free for nowUp to 10 sessions at once
MaxUS$200 a monthOne user; larger weekly quota, no daily capUp to 10 sessions at once
TeamsFrom US$80 a monthUp to 200 users; full seats US$40 a month each, flex seats free on shared creditsUnlimited sessions
EnterpriseOn requestBilled in ACUs; SSO, admin controls, dedicated deploymentUnlimited sessions

Each paid plan has a usage allowance that refreshes daily and weekly; Max has only a weekly one. Past it you buy on-demand credits, prepaid and charged at the token price of the model you use; they never expire and can top up automatically. How much a quota covers is not published: the pricing page says only that most users will never hit their limits. On Teams, Devin Review and Automations always draw from the shared credit pool, and when credits run out, automations stop and Review falls back to a plain diff viewer.

Devin's usage docs explain what you pay for. Cloud sessions count the agent's actions plus virtual machine time; Windows sessions use about 9% more than Linux; a session sleeps after 5 minutes idle and costs nothing while asleep or waiting on you. Large enterprise customers can join Cognition's AI Productivity Guarantee: if Cognition's own estimate of the engineering hours Devin delivered falls short of what they consumed, it issues free credits, up to US$10 million.

The models you pick move the bill most. Rates for use beyond the quota, from Devin's model price list:

Model in DevinPer million input / output tokens, beyond the quotaNote
SWE-2 FlashFree until December 4, 2026; then US$0.10 / US$0.50Desktop and CLI only
SWE-2Free in Desktop and CLI through November 30, 2026; then US$3 / US$15Not free in cloud sessions
Kimi K3.1 ProUS$2 / US$10Moonshot's model, in four effort levels
GPT-6.1 SolUS$2 / US$10Can be billed to a ChatGPT plan instead
Claude Opus 5.5US$4 / US$20Fast variants cost twice as much
Claude Fable 5.1US$10 / US$50Anthropic's top model

The price list has more than 250 model and effort options. For SWE-2 Flash, Cognition's launch post says "Free through 11/30", four days earlier than the price list's December 4 at 18:00 UTC.

Refunds are tight. The terms updated June 30, 2026 call payments non-refundable unless the terms or the law say otherwise, and Devin's docs offer a refund only before a subscription is used. There is no free trial of a paid plan. Windsurf subscribers who had a US$15 Pro plan keep that price after the switch to quotas in March 2026.

What public tests and users report

Third-party tests.

  • Answer.AI, January 2025. Hamel Husain, Isaac Flath and Johno Whitaker gave Devin 20 tasks over a month: 3 successes, 14 failures, 3 inconclusive. Devin handled glue code, such as pulling a Notion database into Google Sheets, but produced overcomplicated code for an API integration, got stuck parsing HTML in a scraping job and spent more than a day on a deployment Railway does not support. Whitaker summed it up: "Tasks it can do are those that are so small and well-defined that I may as well do them myself, faster, my way." This was the 2024 Devin.
  • Internet of Bugs, April 2024. A frame-by-frame video of the launch demo argues Devin did not complete the Upwork job it was shown doing and took at least six hours on work the author redid in about 36 minutes.
  • Embrace The Red, August 2025. Security researcher Johann Rehberger showed that instructions planted in a GitHub issue made Devin download and run malware, and later posts showed it leaking secrets and opening ports to the internet. He reported the flaws in April 2025 and said Cognition had not given a fix timeline 120 days later. Devin now has security profiles that restrict network, MCP and git access.

A competitor's test. Tenki, which sells its own reviewer, ran seven tools on 50 pull requests holding 122 real bugs in April 2026. Devin caught 44 (36.1%) at 47.3% precision, an F1 score of 40.9, second to Tenki's 41.7 and ahead of Cursor (39.4), CodeRabbit (26.7), Greptile (22.1) and GitHub Copilot (21.4). That run predates Devin Review 2.0.

Cognition's own numbers. Cognition reported in November 2025 that 67% of Devin's pull requests were merged, up from 34% a year earlier, and described Devin as senior-level at understanding a codebase but junior at execution. Its own docs say Devin is most likely to succeed on tasks you could do in three hours or less, with clear success criteria.

User reports. In the Hacker News thread on SWE-2 in September 2026, opinions split. Critics called the CLI buggy, said the web and desktop clients run different harnesses of uneven quality, and objected that SWE-2 can only be used inside Devin. Supporters, including a heavy team user, called Devin Cloud one of the best team-managed cloud agents for its sandboxed testing and API, and one Pro subscriber found SWE-2 Max clearly better than Kimi K3. A separate commenter, who builds a competing product, wrote that Devin Automations has solid UX "but can get very expensive to run".

Reliability. Devin's status page lists 25 incidents from May 13 to November 4, 2026, including Devin Cloud being unavailable on August 13 and the web app going down on September 24.

Privacy, security and where your code runs

Training is on by default. Section 3.3.1 of the terms lets Cognition use customer data to train models and improve its services; a paid plan can opt out on the Data Controls page, after which Cognition says the data is not used and zero data retention is switched on with its model providers. On Teams only an admin can opt out, the Free plan cannot, and Enterprise customers are promised no training without written consent. Outputs belong to you.

Cognition says it has held SOC 2 Type II since September 2024 and works to ISO 27001, and the Windsurf extensions received FedRAMP High authorization in March 2025. The deployment docs say each session runs on its own isolated machine in Cognition's cloud on AWS. Enterprises can move Devin's workspace into a single-tenant network reached over AWS PrivateLink, or run sessions on their own machines with Outposts. Devin's "brain", the service that powers its reasoning, always stays in Cognition's cloud. The privacy policy, updated October 26, 2026, says Cognition shares identifiers with advertising partners in ways some US states treat as a sale, with an opt-out page.

Who should not use Devin

  • Anyone who needs a fixed monthly cost. Quotas are not published in tokens or tasks, and use beyond them is billed per token.
  • Teams that cannot write tight specs. Devin's own guidance asks for clear success criteria, and Answer.AI could not predict which tasks would succeed.
  • Free-plan users with private code they do not want used for training. Only paid plans can opt out.
  • Anyone pointing an unattended agent at untrusted issues or web pages without a network allowlist. The 2025 prompt-injection research shows what can happen.
  • Developers who only want an assistant in the editor. Cheaper tools cover that, and Devin's strengths are in the cloud agent.

Alternatives

OptionWhat it isEntry price, checked November 6, 2026Better choice when
DevinCloud agent, desktop editor, CLI and PR review in one planFree; Pro US$20 a monthYou want to assign tickets to an agent that tests on its own machine
Claude CodeAnthropic's coding agent, in every paid Claude planClaude Pro US$20 a month (US$17 billed yearly)You drive the agent yourself and already use Claude
OpenAI CodexOpenAI's agent on the web, in the CLI, IDE and iOS, with cloud code reviewFree and Go (US$8) with GPT-6 Luna; Plus US$20You already pay for ChatGPT
CursorAI editor with cloud agents and automations; bought by SpaceX in August 2026, per Tech Funding News and Channel InsiderPremium US$30 a month for cloud agents; Plus US$8 runs only Grok modelsYou live in an editor and want agents as an add-on
GitHub CopilotGitHub's agent, cloud tasks and code review, plus Claude Code and Codex as agentsPro US$10 a monthYour code is on GitHub and you want the cheapest cloud agent
FactoryDroid agents on desktop, CLI and in the cloudPro US$20 a month; Teams US$60 plus US$40 a seatYou want a similar agent with a different harness

Several of these bundle an agent into a plan you may already have. Devin's edge is a managed cloud computer per task that tests its own work, hand-offs from Slack, Linear and Jira, and enterprise deployment options; its weak spots against them are unpublished quotas and thin outside testing. On GenGrades, T3 Code is a free app that runs the agents you already pay for side by side, much like Devin Desktop's command center; ZCode is Zhipu's free desktop coding agent on its GLM models; Command Code is another terminal agent; and OpenCode Zen sells models for the open-source OpenCode agent. If you want an agent for research or slides rather than code, Manus is the closer fit.

Before committing a team, run the same ticket through Devin and one alternative and compare the merged result and the cost.

Sources

Sources

  1. Devin homepage (Cognition AI, Inc., checked November 6, 2026) (devin.ai), accessed
  2. Devin llms.txt: product surfaces (Cognition, checked November 6, 2026) (devin.ai), accessed
  3. Devin plans and pricing, with the plan comparison table (Cognition, checked November 6, 2026) (devin.ai), accessed
  4. Download Devin Desktop, JetBrains plugin and CLI (Cognition, checked November 6, 2026) (devin.ai), accessed
  5. Join the waitlist for Devin Mobile (Cognition, checked November 6, 2026) (devin.ai), accessed
  6. The AI Productivity Guarantee (Cognition, checked November 6, 2026) (devin.ai), accessed
  7. Devin status and incident history (Cognition, checked November 6, 2026) (devinstatus.com), accessed
  8. Devin Review 2.0 (Cognition, November 5, 2026) (devin.ai), accessed
  9. Windsurf is now Devin Desktop (Cognition, June 2, 2026) (devin.ai), accessed
  10. Sign in with ChatGPT (Cognition, September 29, 2026) (devin.ai), accessed
  11. Devin Security 2.0: Built for the new threat landscape (Cognition, October 26, 2026) (devin.ai), accessed
  12. Introducing Devin Oncall (Cognition, October 15, 2026) (devin.ai), accessed
  13. Hierarchical budgets (Cognition, November 2, 2026) (devin.ai), accessed
  14. Windsurf Extensions are now FedRAMP High authorized (Cognition, March 10, 2025) (devin.ai), accessed
  15. Introducing SWE-2: Pushing the Pareto Frontier (Cognition, September 10, 2026) (cognition.com), accessed
  16. Introducing SWE-2 Flash (Cognition, November 4, 2026) (cognition.com), accessed
  17. New self-serve plans for Devin (Cognition, April 14, 2026) (cognition.com), accessed
  18. Introducing Devin, the first AI software engineer (Cognition, Scott Wu, March 12, 2024) (cognition.com), accessed
  19. Introducing Devin 2.2 (Cognition, February 24, 2026) (cognition.com), accessed
  20. Cognition's acquisition of Windsurf (Cognition, Scott Wu, July 14, 2025) (cognition.com), accessed
  21. More Devins in More Places: Series D (Cognition, May 27, 2026) (cognition.com), accessed
  22. Do it all with Devin: Announcing our Series E (Cognition, September 8, 2026) (cognition.com), accessed
  23. Cognition crosses $1B in annualized revenue run rate (Cognition, September 25, 2026) (cognition.com), accessed
  24. Devin's 2025 Performance Review (Cognition, November 14, 2025) (cognition.com), accessed
  25. Devin Fusion: Frontier Performance at 60% Lower Cost (Cognition, June 29, 2026) (cognition.com), accessed
  26. Cognition Platform Terms of Service (Cognition AI, Inc., last updated June 30, 2026) (cognition.com), accessed
  27. Cognition Privacy Policy (Cognition AI, Inc., last updated October 26, 2026) (cognition.com), accessed
  28. Self-serve plans: seats, Teams minimum, quota and on-demand credits (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  29. Usage: how Devin meters work (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  30. Enterprise billing in ACUs (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  31. Plans and Usage: trials, refunds and payment (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  32. Quota-based usage (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  33. AI Models and token prices, as of November 5, 2026 (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  34. Security at Cognition (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  35. Security Profiles (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  36. Enterprise Deployment (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  37. Computer Use (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  38. Devin CLI (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  39. When to Use Devin (Devin docs, checked November 6, 2026) (docs.devin.ai), accessed
  40. Thoughts On A Month With Devin (Answer.AI, Hamel Husain, Isaac Flath and Johno Whitaker, January 8, 2025) (answer.ai), accessed
  41. Debunking Devin: "First AI Software Engineer" Upwork lie exposed! (Internet of Bugs, YouTube, April 9, 2024) (youtube.com), accessed
  42. I Spent $500 To Test Devin AI For Prompt Injection So That You Don't Have To (Embrace The Red, Johann Rehberger, August 6, 2025) (embracethered.com), accessed
  43. How Devin AI Can Leak Your Secrets via Multiple Means (Embrace The Red, August 7, 2025) (embracethered.com), accessed
  44. AI Kill Chain in Action: Devin AI Exposes Ports to the Internet with Prompt Injection (Embrace The Red, August 8, 2025) (embracethered.com), accessed
  45. AI Code Review Benchmark 2026 (Tenki, a competing code reviewer; last updated May 20, 2026) (tenki.cloud), accessed
  46. Cognition launches new SWE-2 model: discussion (Hacker News, September 10 to 16, 2026) (news.ycombinator.com), accessed
  47. Comment on Devin Automations in 'Our decision on Cursor following its acquisition by SpaceX' (Hacker News, August 29, 2026) (news.ycombinator.com), accessed
  48. Cognition, maker of the AI coding agent Devin, acquires Windsurf (TechCrunch, July 14, 2025) (techcrunch.com), accessed
  49. Windsurf Joins Cognition After Google Hires CEO in $2.4B Deal (Built In, July 16, 2025) (builtin.com), accessed
  50. Cognition raises $2B Series E at $48B valuation to scale Devin (Dealroom, September 2026) (dealroom.co), accessed
  51. Devin maker Cognition raises over $2B at $48B valuation (Tech Funding News, September 9, 2026) (techfundingnews.com), accessed
  52. Cognition Hits $48B Valuation as Demand for Devin Surges (Channel Insider, September 2026) (channelinsider.com), accessed
  53. Claude plans and pricing (Anthropic, checked November 6, 2026) (claude.com), accessed
  54. Codex pricing (OpenAI, checked November 6, 2026) (developers.openai.com), accessed
  55. Cursor pricing (Cursor, checked November 6, 2026) (cursor.com), accessed
  56. GitHub Copilot plans (GitHub, checked November 6, 2026) (github.com), accessed
  57. Factory pricing plans (Factory, checked November 6, 2026) (factory.ai), accessed

Changelog

Changelog

  1. Page created from Devin's and Cognition's own pages, documentation and terms, public tests by Answer.AI, Internet of Bugs, Embrace The Red and Tenki, Hacker News user reports, news reporting and rivals' pricing pages.

Recommended from GenGrades

  • Model review

    Claude Fable 5.2

    Claude Fable 5.2 is Anthropic's most capable public model, released November 9, 2026 at 8 and 40 USD per million input and output tokens, 20% below Fable 5.1. Against Opus 5.5 it costs double per token and, at max effort, scores 65 to 59 on Artificial Analysis's index for 26.54 USD per task against 10.04; Anthropic still recommends Opus first.

  • Model review

    LongCat-2.5

    LongCat-2.5 is Meituan's 1.6-trillion-parameter model for long coding and agent tasks, released on November 2, 2026; unlike LongCat-2.0, it also reads images and video. Its API costs 0.30 USD per million input tokens and 1.20 USD per million output at a launch discount. The weights are not out yet, and every benchmark so far is Meituan's own.

  • Tool review

    T3 Code

    T3 Code is a free, MIT-licensed app from T3 Tools Inc., credited to Theo Browne's team, that runs the coding agents you already pay for, such as Claude Code, Codex, Cursor and OpenCode, side by side from a desktop, web or phone interface. First released on March 7, 2026, its GitHub repository had 28.1k stars on October 17, 2026.

  • Tool review

    Command Code

    Command Code is a terminal coding agent from Langbase, Inc., trading as Command Code and led by Ahmad Awais, that says it learns your coding "taste" from the changes you accept, reject and edit. Announced with 5 million USD in seed funding on February 25, 2026, it now focuses on low-cost open models, with plans from 1 USD a month.

  • Tool review

    ZCode

    ZCode is Zhipu AI's (Z.ai) free desktop coding-agent app for macOS, Windows and Linux, built around its GLM-5.3 model, with model use billed through the GLM Coding Plan from 18 USD a month. First released as Z Code in December 2025, it switched to Zhipu's own agent in June 2026; Z.ai said in August it had passed one million users.

  • Guide

    GPT Image 2 API Price per Image: 23 Providers Compared

    APIMODELS, our own product, publishes a 30-day success rate on every model page, 99.9% for gpt-image-2; of the other 22 GPT Image 2 API providers compared, only OpenRouter and CometAPI publish any model-level reliability figure. Prices were read on November 11, 2026, and the lowest medium 1K price is APIMODELS at 0.025 USD.