---
type: analysis
slug: pro-200-loses-its-usage-per-dollar-edge-over-plus-and-pro-500-sells-speed
title: "Pro 200 loses its usage-per-dollar edge over Plus, and Pro 500 sells speed rather than cheaper compute"
description: "Pro 200's included Work and Codex usage drops from 20x to 10x the Plus allowance. If Plus still costs $20, which the sources do not confirm, Pro 200 now buys the same usage per dollar as Plus. Pro 500 does not restore the discount. It adds Ultrafast, which OpenAI says draws on included usage. Our reading is that ChatGPT plans are starting to work like compute balances. That is our interpretation, not something OpenAI has said."
published_at: 2026-10-26T20:00:00.000Z
topics: ["openai","codex","agents"]
source_urls: ["https://www.infoq.com/news/2026/10/openai-devday-2026","https://openai.com/index/uber-chatgpt-work-rollout","https://community.openai.com/t/fresh-dots-updates-create-in-the-mobile-app-delegate-to-codex/1404834","https://thenextweb.com/news/openai-devday-pro-200-usage-cut-pro-500-plan","https://openai.com/index/devday-2026-recap"]
author: "Andrey Reshetnikov"
author_profile: https://newruntime.com/owner-profile.md
---
# Pro 200 loses its usage-per-dollar edge over Plus, and Pro 500 sells speed rather than cheaper compute

Pro 200's included Work and Codex usage drops from 20x to 10x the Plus allowance. If Plus still costs $20, which the sources do not confirm, Pro 200 now buys the same usage per dollar as Plus. Pro 500 does not restore the discount. It adds Ultrafast, which OpenAI says draws on included usage. Our reading is that ChatGPT plans are starting to work like compute balances. That is our interpretation, not something OpenAI has said.

## Summary

At DevDay 2026 OpenAI released GPT-6.1 Sol. It says Sol comes close to GPT-6 Astra on several evaluations at one-fifth of Astra's token prices. OpenAI also added computer use to a managed Agents API and gave Codex cloud environments. On pricing, Pro 200's Work and Codex allowance falls on October 30 from 20x to 10x Plus, and a $500 Pro tier adds Ultrafast speed. If Plus is $20, the heaviest individual users lose their volume discount. What they can now pay extra for is speed and premium model access.

## What happened

InfoQ reports that the Agents API can now operate graphical interfaces. It also gains multi-agent features from Codex, tool search, tool calling and context compaction, all running on infrastructure OpenAI manages. GPT-6.1 Sol is aimed at coding, computer use and professional tasks. OpenAI prices it at one-fifth of Astra's standard input and output rates, with cached input at $0.10 per million tokens. It has not published how the evaluations were run. Codex added cloud environments, voice input in the CLI and an /agents interface.

According to The Next Web, Pro 500 costs $500 a month and is the only tier with Ultrafast. OpenAI says Ultrafast reaches up to 300 tokens per second in Codex, up to eight times standard speed. Pro 200 is open to new subscribers again. From October 30, its included Work and Codex usage falls from 20x to 10x the Plus allowance, and its weekly GPT-6 Pro message cap falls from 200 to 100. Existing subscribers keep their current limits until October 29 and receive a one-time $2,500 usage credit that expires at year-end. The five-hour limit will not return.

OpenAI's recap also announces an OpenAI Marketplace. Eligible enterprises can apply part of an existing OpenAI commitment to approved partner software. Among the first 32 partners are Figma, Adobe, Sierra, Decagon, HubSpot and Salesforce. Separately, Uber is extending ChatGPT Work from a 1,000-person pilot to more than 30,000 employees, with Sol as the default model. Dots can now be created on iOS and Android, can start or continue Codex threads, and can edit recurring Scheduled Tasks.

## What the archive adds

The archive mostly repeats the four stories in the pack, so the useful additions come from the dossier. OpenAI's Help Center page on Pro tiers confirms that Pro 500 costs $500 a month and includes Astra Ultrafast. It also confirms that Pro 200 stays at $200 with a lower included allowance for new subscribers (d2). The Help Center page on ChatGPT Work and Codex separates conversational Chat from Work and Codex, which it describes as longer, multi-step and software-development work. It says Ultrafast consumes included Pro 500 usage (d3). Our copy of that page is truncated, so we cannot confirm the exact rate at which Ultrafast draws the allowance down. An Epoch AI analysis in our archive estimates how many AI agents could run on available compute and examines agents' context sizes and API costs (d1). Its framing supports one point here: the agent task, not the chat message, is becoming the unit that gets paid for. Our excerpt of that analysis does not include its per-agent cost figures, so we do not cite any.

## Trend signal

Chat messages are short and cost little to serve. Work and Codex tasks are long and multi-step, and more and more of them run without anyone watching. Dots hands them off from a phone, Scheduled Tasks repeat them, and the Agents API runs them on OpenAI's machines. Over the same period, the way OpenAI describes its plans has moved toward measured quantities. Allowances are stated as multiples of Plus, Ultrafast draws on included usage (d3), the retention credit is denominated in dollars, and enterprise commitments can now be spent on partner software. Our inference is that OpenAI is moving its plans toward consumption-based pricing. The evidence shows this happening in several separate places. It does not show a single unified compute currency, and OpenAI has not said it is building one.

## Why now

1) Agent tasks are measured differently from chat. OpenAI's Help Center treats Work and Codex as longer, multi-step jobs that draw on a separate included allowance (d3), and Epoch AI analyzes compute and API cost per agent rather than per message (d1). 2) Speed is now part of the price. Ultrafast is limited to Pro 500 and consumes included usage (d2, d3). 3) The Sol pricing, at one-fifth of Astra's token rates according to InfoQ, makes it possible to steer volume toward a cheaper model. 4) OpenAI has not said why it cut Pro 200 at the same time as cutting per-token prices. Capacity pressure and deliberate repricing are competing hypotheses, and the evidence does not settle which one applies.

## Connections to other events

The arithmetic below assumes Plus still costs $20 a month. The pack does not give its price. Under that assumption, Pro 200 used to buy 20x Plus usage for 10x the price, roughly twice the Work and Codex usage per dollar. From October 30 it buys 10x for 10x, the same rate as Plus. What the extra $180 still buys is GPT-6 Pro messages, now 100 a week, and the convenience of one bill instead of several Plus accounts.

Pro 500 does not bring back cheaper usage. Its distinguishing feature is Ultrafast, up to 300 tokens per second in Codex. According to the Help Center, Ultrafast consumes included usage (d3), so a user who keeps it on spends the allowance faster. In practice, the premium is paid for lower latency, not for more work done. Nothing in our evidence says Pro 500 comes with a larger allowance per dollar.

The pricing points the other way at the model level. If a task needs a similar number of tokens and retries on Sol as on Astra, it should cost about one-fifth as much at list prices. Uber making Sol the default for 30,000 seats fits that logic, although neither company has disclosed contract terms.

There is a tension in these moves. Cheaper per-token prices on Sol and smaller included allowances on Pro 200 happened at the same time, and OpenAI has not said why it cut the allowance. We see two competing explanations. The first is capacity: agent workloads, especially Ultrafast and cloud-executed tasks, may be straining GPU supply, and cutting the heaviest flat-rate users frees capacity. That would fit a cut that is later partly reversed. The second is repricing: OpenAI may be deliberately moving heavy users toward metered API spending or the higher tier. That would fit the dollar-denominated credit and the decision not to restore the five-hour limit. The $2,500 grant equals 12.5 months of Pro 200 fees and expires in about two months. That looks like a retention measure, but we do not know what it costs OpenAI to honor.

## Why it matters

For individual developers, the useful metric is now cost per completed task. With Pro 200 at the same usage per dollar as Plus (assuming Plus is $20), Pro 200 is worth it if you need GPT-6 Pro messages or a single account. For automated, repeating jobs, metered calls to Sol through the API may cost less and be easier to predict. That holds only if Sol actually finishes your tasks without many retries.

For teams building on the Agents API, managed execution removes infrastructure work. It also adds a dependency on OpenAI's runtime, and Pro 200 has just shown that OpenAI can reprice included usage at short notice.

For enterprise buyers, the Marketplace makes an OpenAI commitment usable across partners like Salesforce and Figma. That makes the commitment more flexible and also concentrates more spending with one vendor. Procurement should ask how that commitment is drawn down, not only how many seats it covers.

## What happens next

- **OpenAI will partially restore Pro 200's included Work/Codex allowance above 10x Plus.** (6 months, 20%). Watch for: A Help Center or pricing-page revision raising the Pro 200 multiple. The probability is low because OpenAI explicitly refused to restore the five-hour limit, which points to a deliberate repricing.
- **Anthropic will introduce an individual Claude plan priced above its current highest individual tier.** (6 months, 35%). Watch for: A new tier on Anthropic's pricing page above its current top plan. The probability is moderate because a $500 OpenAI tier creates room to price above it, but competitors may prefer to win over Pro 200 users who leave.
- **Uber or OpenAI will publish at least one quantitative adoption or usage metric for the 30,000-employee ChatGPT Work rollout.** (12 months, 60%). Watch for: An OpenAI customer story or Uber post giving active-user share or task counts. The companies have said they will share adoption and outcome information, but such promises often turn into qualitative case studies.

## Limits and uncertainty

The Plus price is assumed, not sourced. If it has changed, the per-dollar comparison changes with it. We do not know how large the Plus allowance is in absolute terms, so "10x Plus" cannot be converted into tokens or tasks. We cannot confirm from our truncated source how fast Ultrafast drains the allowance. OpenAI has not published the methodology behind the Sol vs Astra comparison. If Sol needs longer contexts or more retries, it could lose much of its fivefold price advantage. We have no independent cost-per-task figures for Sol. Uber's terms, usage and outcomes are undisclosed. The reasons we give for the cut and the credit are hypotheses, not anything OpenAI has said.

## Skill to build now

Measure cost per completed agent task. As a first exercise, pick five repeating Codex or Work tasks and run each on Sol through the API. For each run, log input, cached and output tokens, the number of retries, and whether the task succeeded. Compute dollars per successful task and compare it with how much of your plan's included usage the same tasks consume.

## Reader takeaway

Before October 30, take one week of real Codex or Work tasks and log how much of your included usage they consume on your current plan. Then rerun a sample through the API on Sol and record dollars per successfully completed task. Move scheduled and repeating jobs to whichever option is cheaper per finished task. Keep the subscription, and Ultrafast if you have it, for interactive work where waiting is what actually costs you.

## Open question

Did OpenAI cut Pro 200 because agent workloads are straining compute capacity, or because it wants heavy individual users on metered API billing? A later partial restoration would support the first explanation. A move to dollar-denominated usage purchases would support the second.

## Sources

- [OpenAI DevDay 2026 Recap for Developers](https://www.infoq.com/news/2026/10/openai-devday-2026)

- [Uber brings ChatGPT Work to all employees with OpenAI](https://openai.com/index/uber-chatgpt-work-rollout)

- [Fresh dots updates: Create in the mobile app, delegate to Codex](https://community.openai.com/t/fresh-dots-updates-create-in-the-mobile-app-delegate-to-codex/1404834)

- [OpenAI halves Pro 200 usage and launches a $500 ChatGPT plan at DevDay](https://thenextweb.com/news/openai-devday-pro-200-usage-cut-pro-500-plan)

- [DevDay 2026 Recap | OpenAI](https://openai.com/index/devday-2026-recap)

- [About ChatGPT Pro tiers](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers)

- [ChatGPT Work and Codex](https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex)

- [Introducing GPT-6 Sol and Luna](https://openai.com/index/introducing-gpt-6-sol-and-luna/)

- [GPT-6.1 Sol (Max) Intelligence, Performance & Price Analysis — Artificial Analysis](https://artificialanalysis.ai/models/gpt-6-1-sol/)

- [GPT-6.1 Sol Model — OpenAI API](https://developers.openai.com/api/docs/models/gpt-6.1-sol)

- [Plans & pricing — Claude by Anthropic](https://claude.com/pricing)

- [Uber brings ChatGPT Work to all employees with OpenAI](https://www.uber.com/gp/en/blog/uber-announce-work/)

- [Enterprise signals: What frontier firms are doing differently](https://openai.com/signals/enterprise-data/)

- [Analytics API](https://learn.chatgpt.com/docs/enterprise/analytics-api)

Author: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Openai](https://newruntime.com/topics/openai.md): Explore the openai topic hub.
- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- topic hub: [Agents](https://newruntime.com/topics/agents.md): Explore the agents topic hub.
- collection: [Analysis](https://newruntime.com/analysis.json): Open the latest authorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)