vibeplan.cc Open the interactive comparison

AI Coding Plan Comparison 2026

Which AI coding subscription gives you the most work per dollar? This AI coding plan comparison ranks 80 plans by tokens and requests per dollar paid, using only published prices and quotas, with every number rebuilt daily from 30 official sources. Prices are what you pay, fees included.

What one dollar buys

Median across the models each plan covers, cache-aware and normalised per dollar paid. A plan enters this comparison when at least three of its models have a published rate; plans without API access or without a published included amount are listed separately on the site instead of being mixed in.

#	Plan	Price / mo	Tokens / $	Requests / $	Models	Quota
1	OpenDesign Goopendesign	$10.00	372.5M	4,837	10	disclosed
2	OpenCode Goopencode	$10.00	90.9M	1,604	35	disclosed
3	OpenCode Go Plusopencode	$40.00	68.2M	1,203	35	disclosed
4	Claude Max 20xanthropic	$200.00	42.2M	828	5	measured
5	Ollama Cloud Proollama	$20.00	39.2M	770	15	disclosed
6	Ollama Cloud Maxollama	$100.00	39.2M	770	15	disclosed
7	Ollama Cloud Teamollama	$500.00	26.2M	513	15	disclosed
8	Command Code GOATcommand-code	$10.78	18.0M	327	88	disclosed
9	Kimi Code Vivacemoonshot	$199.00	13.4M	238	4	disclosed
10	Kimi Code Allegromoonshot	$99.00	13.2M	235	4	disclosed
11	Command Code Procommand-code	$21.56	11.9M	234	88	disclosed
12	Kimi Code Moderatomoonshot	$19.00	9.7M	174	4	disclosed
13	Kimi Code Allegrettomoonshot	$39.00	9.5M	170	4	disclosed
14	Cursor Ultracursor	$200.00	7.5M	147	11	reported
15	GitHub Copilot Maxgithub	$100.00	7.3M	144	35	disclosed
16	GitHub Copilot Pro+github	$39.00	6.6M	129	35	disclosed
17	GitHub Copilot Progithub	$10.00	5.5M	108	35	disclosed
18	JetBrains AI Ultimatejetbrains	$30.00	4.4M	86	3	disclosed
19	Command Code Max 10×command-code	$107.80	3.9M	70	88	disclosed
20	Command Code Max 20×command-code	$215.60	3.9M	70	88	disclosed

Snapshot 2026-10-02. Rates move with provider prices and model lineups, so treat this table as today's answer, not a permanent one. 51 of 80 plans publish an included amount; the rest stay marked undisclosed.

How the numbers are computed

Questions people ask

What is an AI coding plan?

A fixed monthly subscription that gives you an AI coding agent plus a quota of model usage, instead of paying per token. This comparison covers 80 plans from 30 official sources: prices, published quotas, per-model rates and what one dollar actually buys.

Which coding plan gives the most tokens per dollar right now?

On 2026-10-02, OpenDesign Go at $10.00 per month leads the ranking with a median of 372.5M tokens per dollar across 10 comparable models (median 4,837 requests per dollar). The interactive chart shows the whole Pareto front of AI score against tokens per dollar, so you can trade quality against volume instead of chasing a single number.

How is the comparison calculated?

Every plan is converted to requests and tokens per dollar paid, using the model's published API prices: cost per request = (0.05 x input + 0.95 x cached write) x input tokens + cached read x cached read tokens + output x output tokens, per million. Cash credits convert at the published rate, weekly windows scale x4.33 to a month, and five-hour windows are throughput caps, never multiplied into a monthly volume.

Do these plans publish how much usage is included?

No, not all of them. 51 of 80 plans in the catalogue publish an included amount; the rest only show it at checkout or not at all. Undisclosed stays undisclosed here: those plans appear with their list price and a note instead of a guessed number, so a made-up quota never moves the ranking.

Which plans keep my code out of training, and which keep zero retention?

Training and retention are tracked per model and vendor, not per brand: 4 model rows in the catalogue are flagged as training on your data and are hidden by default. Zero retention (confirmed zero-day retention for that exact model at that vendor) applies to 54 model rows. A vendor saying "zero retention" on its pricing page is not enough; the statement has to cover the model you actually use.

Why is this coding plan comparison built on tokens per dollar instead of tokens per plan?

Because "60 for 10" tells you nothing: a plan with a large credit number can buy less work than a cheaper plan with better cache pricing. Rates here are normalised per dollar paid, cache-aware, and computed per model, so plans with different meters become comparable. Plans that cannot be compared that way (no published amount, no API access) are listed separately rather than mixed into the ranking.

How often is the data updated?

Daily. A scheduled build fetches all 30 sources, parses them, and rebuilds the dataset; the changelog is generated from the repository commits, so every change is traceable to a commit. The dataset is served as JSON at https://vibeplan.cc/data/latest.json and each plan row keeps its verification date.

Can I reuse the numbers?

Yes. The data is MIT licensed and served as JSON. Link to the page rather than copying numbers, because rates change daily as prices and model lineups move.

Data, sources and license

Every source is listed with its URL and parser in sources.yml, manual entries and their verification dates in overrides.yml. The dataset is MIT licensed: reuse it, but link rather than copy, because it changes daily.

Interactive chart and table · latest.json · Atom feed of changes · Repository