Devin Pro: $20/month

Devin Pro covers about 234 turns of balanced coding work a month, 12 per dollar. That is the 16th most of the 37 plans with a usage figure. At $20 or less, Claude Pro covers the most: about 1.1k.

What it includes

Price	$20/month
Coding agent	Devin CLI and Desktop
Limit, as Cognition words it	A daily and a weekly quota, drawn down at API prices. No size is published.
Usage at API prices	$128–$147 a month, 6.4–7.3× the price
Evidence	Measured · checked Nov 12, 2026
Strongest model	Fable 5.2, 64.6 on the Intelligence Index

Models

Claude Fable 5.2 · up to 64.6
Claude Opus 5.5 · up to 59.1
Claude Sonnet 5.5 · up to 54.5
Claude Haiku 5.5 · up to 36.5
GPT-6 Astra · up to 56.9
Muse Wonder · up to 43.6
Gemini 4 Argon · up to 50.4
Gemini 3.8 Flash · up to 30.8

Value by kind of work

Turns of mixed coding work the allowance covers in a month, each kind on the plan’s cheapest model that is good enough. How this is worked out.

Work mix	Turns a month	Per dollar
Light only (100% light)	3.3k–3.8k	164–188
Mostly light (40% normal, 60% light)	575–660	29–33
Balanced (5% frontier, 15% hard, 60% normal, 20% light)	218–250	11–13
Mostly hard (10% frontier, 40% hard, 40% normal, 10% light)	146–167	7.3–8.4

On the balanced mix: frontier work on Fable 5.2 (max reasoning); hard work on Opus 5.5 (xhigh reasoning); normal work on Gemini 4 Argon (high reasoning); light work on Haiku 5.5 (high reasoning).

How the figure was obtained

Measured: sessions on four different models read against the weekly quota (Real API Pricing, October–November 2026) came to list-priced usage worth $29–34 a week. Fable 5.2 is assumed to draw like the Fable 5.1 that was measured. GPT usage can be billed to a ChatGPT subscription instead.

Compare

Other Devin tiers: Devin Max ($200)

At a similar price: Claude Team Standard ($20, about 1.3k turns) · Claude Pro ($20, about 1.1k turns) · Claude Pro yearly ($16.67, about 1.1k turns) · Google AI Pro ($19.99, about 293 turns)

See every plan on the charts, or the best plan at each budget.