Skip to content
gentic.news — AI News Intelligence Platform
Connecting to the Living Graph…

Devin

product stable
Devin

Autonomous AI software engineer by Cognition. Can plan, code, debug, and deploy entire projects end-to-end using real developer tools.

4Total Mentions
+0.07Sentiment (Neutral)
0.0%Velocity (7d)
Share:
View subgraph
First seen: Apr 4, 2026Last active: Jul 13, 2026
Profile✓ Researched · 2026-06-24high confidence

Devin, built by Cognition AI, is an autonomous AI software engineer that takes a natural-language task and independently plans, writes, tests, debugs, and ships code in its own cloud sandbox (with a browser, shell, and editor), then opens a pull request. Marketed as a teammate you delegate work to rather than an autocomplete tool, it now anchors a full product suite after Cognition acquired Windsurf in July 2025 and rebranded that IDE to "Devin Desktop" in June 2026. Cognition raised over $1B at a ~$25B pre-money valuation in May 2026 and reports that 89% of its own internal code is written by Devin.

Built by
Cognition AI
Category
Autonomous SWE agent (cloud-based AI software engineer), now paired with an agentic IDE (Devin Desktop, formerly Windsurf)
Open source
No
Models
Devin originally ran on third-party frontier models (e.g., Anthropic Claude and OpenAI GPT). Cognition now also ships its own proprietary agent-optimized models — SWE-1.5 (released Oct 2025) and SWE-1.6 (late 2025/2026) — served via Cerebras at up to ~950 tokens/sec. Through the Agent Client Protocol in Devin Desktop, other agents (Claude Agent, OpenAI Codex, Gemini CLI, Junie, OpenCode) can also run inside the editor. Cognition does not fully disclose which model powers Devin's autonomous cloud agent at any given time.

Available on

Web app (Devin Cloud dashboard)Slack integrationGitHub / GitLab (PR workflow)Devin APIDevin Desktop IDE (formerly Windsurf; VS Code-based, macOS/Windows/Linux)IDE extensions for VS Code and JetBrains

Pricing

Core: $20/month entry with pay-as-you-go Agent Compute Units (ACUs) billed at ~$2.25/ACU (1 ACU ≈ 15 min of active work); capped at ~10 concurrent sessions. Team: $500/month including 250 ACUs at ~$2.00/ACU, unlimited concurrent sessions, Devin API access, and early features. Enterprise: custom pricing with SaaS or VPC deployment for data isolation/compliance. Seats are unlimited across plans — you pay for working time, not licenses. The separately-sold Devin Desktop IDE has its own tiers (a Pro tier around $15/month was introduced in 2026, plus a free tier on the SWE model at lower speed). Note: ACU usage costs can be unpredictable for long/complex tasks.

Benchmarks

  • SWE-bench Verified (Devin 2.0, single-agent pass@1, no human-in-loop)45.8%
  • SWE-bench Lite (original Devin, 2024)13.86%
  • SWE-bench Pro (SWE-1.5 model)~40.1%
  • SWE-bench Pro (SWE-1.6 model)~50.4%

Key capabilities

  • Fully autonomous task execution in a sandboxed cloud workspace with its own shell, browser, and code editor — runs end-to-end with little human input
  • Opens pull requests directly on GitHub/GitLab and integrates with Slack for delegation
  • Parallel/concurrent sessions — fire off many independent tasks at once (unlimited on Team plan)
  • Devin API for embedding the agent into CI/CD and custom automation workflows
  • Codemaps — AI-annotated visual code navigation for monorepos and legacy codebases (introduced via Windsurf integration)
  • Devin Review — automated PR code review (Fast and Comprehensive modes), bundled into paid IDE tiers

Strengths

  • +Genuinely autonomous on well-scoped, parallelizable tasks (bug fixes, migrations, refactors) — you can delegate and walk away
  • +Strong real-world traction with major enterprises (Mercedes-Benz, Goldman Sachs, Santander, NASA) and a complete suite spanning cloud agent + IDE that competitors don't match at scale
  • +Low $20 entry point makes it easy to trial; unlimited seats means you pay for work done, not headcount
  • +Proprietary SWE-1.5/1.6 models are extremely fast (~950 tok/s via Cerebras), reducing agent latency

Limitations

  • Usage-based ACU billing is hard to budget — costs can balloon on long or complex tasks, and Core users can run out of credits mid-task
  • Benchmark scores (45.8% on SWE-bench Verified) trail Claude-based agents at the top of leaderboards; works best on well-scoped tasks and struggles with ambiguous or sprawling ones
  • Rapid product churn — frequent rebrands (Windsurf → Devin Desktop) and deprecations (the Cascade agent is end-of-life July 1, 2026) create migration friction for existing users

Best for

Engineering teams that want to offload well-scoped, parallelizable tasks (bug fixes, migrations, refactors, routine PRs) to an autonomous agent, especially enterprises needing VPC deployment — not a fit for those wanting a cheap, predictable IDE autocomplete.

What's new

June 2026: Cognition rebranded the Windsurf IDE to "Devin Desktop" via an over-the-air update, shifting the default experience to an "Agent Command Center" hub and adopting the open Agent Client Protocol so rival agents (Codex, Claude Agent, Gemini CLI, Junie) run inside one editor. This followed Cognition's >$1B raise at a ~$25B pre-money valuation (May 2026), ~$492M annualized revenue run-rate, and the release of its fast proprietary SWE-1.5/SWE-1.6 models. The legacy Cascade agent reaches end-of-life on July 1, 2026.

Sources (12)
  1. 1. devin.ai
  2. 2. cognition.com
  3. 3. techcrunch.com
  4. 4. techcrunch.com
  5. 5. cognition.ai
  6. 6. cognition.com
  7. 7. cognition.com
  8. 8. cognition.ai
  9. 9. eesel.ai
  10. 10. aitooltier.com
  11. 11. nxcode.io
  12. 12. morphllm.com

Signal Radar

Five-axis snapshot of this entity's footprint

live
MentionsMomentumConnectionsRecencyDiversity
Loading radar…

Mentions × Lab Attention

Weekly mentions (solid) and average article relevance (dotted)

mentionsrelevance
01
Loading timeline…

Timeline

No timeline events recorded yet.

Relationships

5

Competes With

Developed

Benchmarked On

Uses

Frequently appears with

2

Entities that show up in the same articles — shared coverage, not a stated relationship.

Recent Articles

1

Predictions

No predictions linked to this entity.

AI Discoveries

No AI agent discoveries for this entity.

Sentiment History

+10-1
Positive sentiment
Negative sentiment
Range: -1 to +1
WeekAvg SentimentMentions
2026-W29-0.101