Skip to content
Models
PLAN Free $0/mo PLAN Pro $20/mo PLAN Max $200/mo PLAN Teams $80/mo min + $40/seat Cognition FrontierCode (Oct 9) — score · cost per task Claude Opus 5.5 54.6% · $0.80/task Claude Fable 5 53.5% · $13.09/task Claude Opus 5 53.4% · $4.31/task GPT-6 Astra 53.3% · $4.59/task Claude Sonnet 5.5 52.1% · $1.59/task Claude Fable 5.1 50.9% · $3.28/task GPT-6.1 Sol 50.2% · $0.36/task SWE-2 50.0% · $1.18/task GPT-6 Sol 49.3% · $2.07/task Grok 4.6 48.0% · $2.88/task SWE-1.7 42.0% · $1.97/task SWE-1.6 9.4% · $0.49/task Devin CLI: v3000.11.92 Devin Cloud Updates Chat and Session Management Cognition adds Devin to Pylon support platform, offers $15,000 credits Cognition adds Gemini 4 Argon to Devin Desktop and CLI with 75% discount Devin CLI Update Devin Desktop: Continuity — 2026-10-12 Cognition launches FrontierCode 1.1 benchmark to assess AI‑generated p… Devin Cloud: Memory and dreaming · Session resume window is now 14 day… PLAN Free $0/mo PLAN Pro $20/mo PLAN Max $200/mo PLAN Teams $80/mo min + $40/seat Cognition FrontierCode (Oct 9) — score · cost per task Claude Opus 5.5 54.6% · $0.80/task Claude Fable 5 53.5% · $13.09/task Claude Opus 5 53.4% · $4.31/task GPT-6 Astra 53.3% · $4.59/task Claude Sonnet 5.5 52.1% · $1.59/task Claude Fable 5.1 50.9% · $3.28/task GPT-6.1 Sol 50.2% · $0.36/task SWE-2 50.0% · $1.18/task GPT-6 Sol 49.3% · $2.07/task Grok 4.6 48.0% · $2.88/task SWE-1.7 42.0% · $1.97/task SWE-1.6 9.4% · $0.49/task Devin CLI: v3000.11.92 Devin Cloud Updates Chat and Session Management Cognition adds Devin to Pylon support platform, offers $15,000 credits Cognition adds Gemini 4 Argon to Devin Desktop and CLI with 75% discount Devin CLI Update Devin Desktop: Continuity — 2026-10-12 Cognition launches FrontierCode 1.1 benchmark to assess AI‑generated p… Devin Cloud: Memory and dreaming · Session resume window is now 14 day…

Explainers · News

Coming back to Devin Desktop? Adaptive mode is the setting returners keep choosing

Devin Central illustration
Devin Central illustration
Routing isn't just a cloud-agent economics story. It's the default answer to 'which model should I pick?' on the desktop, too.

The most useful product reviews are from people who left and came back — they see the delta, not the baseline. In early August, @tristanbob returned to Devin Desktop after a few months away (still on a grandfathered original Windsurf plan, he notes) and made two observations worth unpacking.

First, his configuration choice: Adaptive mode, “because I want my credits to be used as efficiently as possible.” Second, the result: it “worked great” — two PRs pushed to an open-source project after local testing — with one complaint: the thinking output is incredibly verbose, visibly repeating itself, re-checking rules, correcting itself mid-stream.

What Adaptive actually is

Adaptive has been in the Desktop changelog since v1.9600.38 (April 6, 2026): a model-picker option that “intelligently selects the best model for each task, helping you make your quota last longer by avoiding overuse of premium models,” drawing down quota at a fixed per-token rate, available on Pro, Max, and Teams plans. The same release put token pricing directly in the model picker. Two follow-ups within the week — improved visibility of which model Adaptive chose (v1.9600.40) and a bug fix for the router getting stuck on the first model, with quota reset for affected users (v1.9600.41) — show both the demand and the teething.

That lineage matters because Adaptive is the Desktop-side expression of the same bet Cognition later made loudly in the cloud with Fusion: the user shouldn’t be picking models per task; a router should. Fusion does it mid-session with a frontier lead and a cheap sidekick; Adaptive does it per-task against your quota. Different mechanisms, one thesis — and a returner choosing Adaptive specifically to stretch credits is that thesis working as designed.

The other end of the funnel

For contrast, the same week’s Wire brought a classic first-contact report: @MussaratShams installed Devin Desktop, gave it two prompts, and had a working todo app in minutes. Demo-grade, sure — but the two posts bracket the Desktop’s current audience nicely. New users get the instant-gratification path; returning users find a maturing efficiency layer (Adaptive, plus the 3.7.16-era permission layering and plugin sync) that wasn’t there when they left.

If you’re coming back after months away, the practical takeaway from the field: start in Adaptive, watch which models it picks (the picker now shows you), and judge the verbose thinking for yourself — it’s the one consistent gripe in an otherwise positive returner report.

Method: Adaptive's behavior, availability, and pricing details are from the official Desktop changelog (fetched 2026-08-12). Both user experiences are field reports from X — impressions, not benchmarks.

References

  1. Devin Desktop changelog — Introducing Adaptive (v1.9600.38), visibility (v1.9600.40), router fix (v1.9600.41), v3.7.16 notes.
  2. @tristanbob’s return report — Adaptive for credit efficiency; verbose thinking; two OSS PRs.
  3. @MussaratShams’s first-contact report — two prompts to a working app.
  4. Devin Fusion — the cloud-side routing sibling.

Editorial from Devin Central — a fan news desk, not Cognition. Devin is a trademark of Cognition.