OpenAI dots
OpenAI
As of 9 Oct 2026, PAEval has compiled 31 capability records for OpenAI dots from 11 source URLs; 27 come from the vendor's own documentation and none has yet been independently observed. Assistant Benchmark (updated 8 Oct 2026) gives it 8.4 on its 1–10 scale, the mean of the 7 of 15 dimensions it scored, ranking #3 of 12 scored products. Scores from different evaluators use different scales and are not combined. PAEval does not publish its own score or rank for it.
The primary dot after the release on 2026-09-29, GPT-6 Astra; fixed Pro/Business Premium/Enterprise, permissions and cloud/local/Codex environment; specialist dots preview to create new entities, do not mix ChatGPT ordinary chat/Work/Codex tasks.
- Independent0
- Vendor27
- Third-party report0
- Not established4
31 records · 11 source URLs · 1 external result(s)
Capability summary
Strongest status per domain; every underlying record is listed below. Availability does not establish task success.
Positioning & scope
A primary dot that continues to advance personal responsibilities and projects; can research, produce work, and coordinate parallel tasks.
Regions, eligibility & access
Pro is gradually opened, excluding EEA, Switzerland, and the United Kingdom; Business Premium covers all ChatGPT support areas; Enterprise includes Edu/Healthcare and requires the administrator to enable beta; not available for those under 18 years old.
Create in desktop web/desktop app, Windows also has app; mobile app must have been created and corresponding updates available; mobile web is not supported.
ChatGPT localization/model multi-language does not equal dots language execution coverage; dots independent and complete language list has not been verified.
Pricing & quotas
The first dot is included in Pro, with no independent surcharge; Pro is currently 100/200/500 US dollars per month, and 500 includes Ultrafast, but it cannot be inferred that dot automatically uses this speed.
Business Premium monthly payment is US$125/seat/month, annual payment is equivalent to US$100/seat/month; there are at least two paid seats in the workspace, and Standard/Premium can be mixed.
Limitations: The price is USD for regional reference; do not write the Premium price per seat as the lowest total cost of the new workspace.
Including deep work quota, temporarily increased in the first month of release; additional dots/speed increase/workload are planned in the future.
Limitations: The amount of each dot task/credit during the stable period, the price of additional dots, and refunds for failed work have not been verified. Unlimited execution cannot be promoted.
Autonomy & background work
Active research has read-only tool restrictions; background tasks after authorization can perform actions according to normal approval rules, and the two must be displayed separately.
It supports fixed schedules, available service event triggers, and can also be automatically paused/waked up according to task status to follow up; only connecting services will not automatically create monitoring.
Execution environments
Own continuous cloud computer/browser, independent files, software and website sessions; the user's device can still work in the cloud and take over the browser when the user device is turned off.
You can authorize a personal computer and create a local Work/Codex task; the computer needs to be online and the app is open; it can also be executed in the default Codex cloud environment, the latter does not require the computer to be online.
The mobile app is the access/approval portal, and independent cloud phone or phone native GUI control has not been verified.
Calls, messages & identity
The email address connected to the user can be used; the startup help page clearly states that launch does not have an independent email address; another Slack identity can be set up.
Limitations: A sentence in the Privacy FAQ mentions email identity, which is inconsistent with the boundaries of the start page; independent mailboxes cannot be marked as universally supported.
You can DM/mention dots and authorize interaction with others, and use context across channels but do not automatically copy all messages; entering a channel does not mean automatically turning on monitoring.
The out-of-box third-party phone outbound calling/incoming call answering/meeting joining scope is not checked; calling a user does not mean calling a business.
Connectors & tools
Officially, it can connect to 4000+ applications; plug-in permissions are shared with ChatGPT/Work/Codex, and reading and writing are still authorized by account and plug-in.
Limitations: 4000+ are ecological coverage statements, which are not verified one by one and are not guaranteed to be written in all.
Supports available plug-ins; plug-ins can include MCP services and skills; local skills are connected to the computer and operating environment.
The product API that can directly create/control the primary dot has not been approved; the OpenAI model/Agents API shall not be used to replace this evidence.
Memory & personalization
Continuous communication with ChatGPT memory/recent context; turning off ChatGPT Memory stops sharing but does not delete the received information.
Currently, you cannot view/modify/delete dot memory one by one. Deleting dot will delete its own context; the plug-in will disconnect the existing context.
Limitations: Library files, independent tasks/conversations and ChatGPT memories have their own deletion entries. Deleting dot does not mean clearing everything.
Channels & interfaces
The same dot can chat across ChatGPT, Slack, and Teams; users can initiate real-time voice and send text at the same time; the task continues after the voice ends.
SMS is a limited beta, only available to some US Pro users; Business/Enterprise is not available; message/traffic charges may apply.
The public documents I read clearly indicate that launch cannot proactively call users, and will be rolled out later; do not combine user-initiated voice calls with agent callbacks.
Conditions: The range of publicly available documents that have been read
Limitations: Version is sensitive; it will not be disclosed in the future/the grayscale status of the account is unknown and requires official updates.
Research & deliverables
Cloud research, data analysis, documentation/demo and software construction testing; images/files can be used as input, and products can be extended through plug-ins and tools.
Limitations: The plug-in/environment determines the specific format and multi-modal tools, and all OpenAI model capabilities cannot be regarded as dot products being delivered.
Transactions & bookings
Custom Rules cover sharing, purchasing and access permissions; browsers/plug-ins can perform authorization steps, and sensitive steps need to be completed by the user.
Limitations: This round does not establish merchant coverage/transaction rates or universal wallet purchasing capabilities; reservations/payments must be verified based on actual available tools.
Collaboration & parallelism
A primary dot can dispatch multiple background agents/tasks; subsequent dots and specialist dots are in different stages and cannot be merged into open multiple persistent entities.
Limitations: There are no official values for the concurrency hard limit and shared resource isolation.
Reliability & recovery
Activity displays ongoing/completed tasks, files and pending tasks; supports taking over and changing priorities; task completion does not automatically prove that the results have been delivered.
Pause only stops the main task; assigned tasks and future plans must be stopped separately, and external actions taken will not be rolled back.
Permissions, security & data
Independent auto-review is reviewed according to instructions, permissions and rules; rules cannot break the built-in security requirements, and users need to complete password changes, etc.; rules may still be executed incorrectly.
Business/Enterprise/Edu does not train by default; personal plans are set according to Improve the model for everyone; active research itself does not directly train, and it is processed according to the settings after entering the qualified dialogue/task.
Storage/transmission encryption; turning off training does not exclude limited security manual review; removing the plug-in only stops new access.
External measurements & hands-on results
Different methods, samples and dates. Not combined into one success rate.
ChatGPT Pro, desktop web/Mac/mobile phone/Slack, default Custom Rules ↗
- Evidence type
- independent_small_sample_test
- Tested
- from: 2026-09-30; to: 2026-10-01
- Sample
- tasks: 10; testers: 1
- Reported
- name: task_average; value: 8.8; scale: 0–10; kind: author_scored_tasks; name: overall_experience; value: 9.2; scale: 0–10; kind: subjective_opinion_excluded_from_index
- Findings
- The single scheduled task was successful; the formula was damaged after the spreadsheet was imported into Google Sheets, and mail sorting still requires supervision.
- Limitations
- Not tested for multi-week reliability, memory learning, rule enforcement, Teams or full Slack context.; The same author has different tasks/dates across products, so the same scale cannot be used as a unified review.; Officially stated features are still not fully verified by this review.
Analyst notes
Key differences
- The continuous primary dot superimposes the ChatGPT plug-in ecology and the optional local/save cloud environment, and the measurement needs to record the execution point.
- Clearly distinguish between read-only active research and authorized background actions; Pause does not stop all subtasks and plans.
- Context can cross channels, but the current item-by-item memory control is insufficient, and logging off/disconnecting does not automatically delete existing context.
- The Pro region threshold is different from the Business Premium region rules. The first month's quota cannot be used as long-term pricing; specialist dots are previewed separately.
Known unknowns
- dots independent task language list
- Actively call grayscale range after release
- Standalone mailboxes currently generally available
- Third-party outbound calling/meeting joining
- Native cloud phone
- primary dot public API
- Long-term measurement of deep work/concurrency upper limit/task rollback
Source conflicts and how they were resolved
- topic: Pro pricing; older: Search cache still says Pro200 suspended; newer: Real-time English text column Pro100/200/500 and 200 has been reopened; resolution: Use real-time text instead of caching old summaries.; source ids: chatgpt_pro
- topic: standalone_email; older: The Getting Started Help clarifies that launch does not have a separate email address; newer: Privacy FAQ generally states that even if you have your own email or Slack account; resolution: There is no explicit release/setup document, and unlabeled independent mailboxes are generally available.; source ids: dots_start; dots_privacy
- topic: pause_scope; resolution: The general Pause description is corrected by the refined Controls page: does not automatically stop subtasks/cancel plans.; source ids: dots_start; dots_controls
Detailed capability observations
34 capabilities in 8 groups, reviewed cell by cell. Unknown cells were not researched or lacked sufficient public evidence.
| External contact and communication | ||
|---|---|---|
| User initiates real-time voiceTwo-way real-time voice between the user and his own agent; it does not mean that the agent calls out to a third party. | Documented | Officially supports users to initiate voice calls with dot within ChatGPT. |
| The agent proactively calls the userAgent initiates a call to the user; records in-app calls or phone network. | Unknown | The official document describes that users cannot be actively called when released; whether the current capabilities have changed, the evidence still needs to be updated. |
| Outbound calls to third partiesDial business or personal numbers as authorized and have two-way communication; voice chat and generated audio are not… | Unknown | No official information has been found that is sufficient to confirm that the product can natively make third-party calls; the availability/unavailability of all products cannot be inferred from the in-app voice or this account tool. |
| Answer external callsA third party can dial into a number or call portal controlled by the agent, and the agent will actually answer the… | Unknown | Not researched yet |
| Send and receive messages across channelsActual sending and receiving via email, SMS/RCS, WhatsApp, Slack, Teams, etc.; only drafting, not sending. | Documented | Officially supports ChatGPT, Slack, and Teams contact information; you can communicate with others in Slack with authorization. |
| Participate in live meetingsActually join the meeting and listen, transcribe or speak; recording each operation and reading the meeting minutes is… | Unknown | Not researched yet |
| Agent independent external identityAgents have independent email addresses, numbers or platform identities; they are strictly distinguished from borrowed… | Unknown | The official announcement does not provide an independent email address for agents; the Slack contact portal does not separately prove a complete and independent external identity. |
| Execution environments and devices | ||
| Cloud browser operationThe agent opens web pages, operation forms and cross-site processes in the cloud; ordinary search interfaces are not… | Documented | The official provides its own cloud browser, which can be taken over by users. |
| Cloud computers and terminalsAgents have operational desktops, file systems, and/or terminals; record persistence and isolation units. | Documented | The official provides independent cloud computers for files, research and running software. |
| Cloud Phone and Mobile AppThe agent can operate apps in cloud Android/iOS or equivalent mobile systems; mobile clients and mobile remote control… | Unknown | No evidence was found that the product natively provides a cloud phone that can operate the mobile system/App; the mobile client and cloud computer cannot be used as alternative evidence. |
| User computer operationAccess files or operate applications on authorized user computers; clarify system, online conditions and operating… | Documented | Officially supports connecting to user computers; processing files and applications through separate native tasks. |
| User mobile phone operationReading and writing the system/App or operating UI on the authorized user's mobile phone; only notification and… | Unknown | Not researched yet |
| Application connections and tools | ||
| Connector readRead accounts, mail, calendars, files, etc. through supported connectors; log services and read objects. | Documented | Officially supports connected and authorized plug-ins to read information. |
| Connector writeCreate, modify, or send via a supported connector; having a read connector does not imply write access. | Documented | The official example plug-in can handle document and code changes; the actual write permission of each service needs to be verified separately. |
| Custom connectors and toolsSupport users/developers to access custom APIs, MCPs or plug-ins; record protocols, hosting and permissions. | Unknown | Not researched yet |
| Login and session persistenceYou can securely authorize services and reuse the login for subsequent tasks; it is not equivalent to obtaining the… | Documented | The official provides a private browser login process, and the login session can be continuously used for subsequent tasks. |
| Continuous execution and initiative | ||
| Offline background promotionAfter closing the client, authorized tasks can still continue; the duration and amount are separately noted. | Documented | Official instructions indicate that cloud work can continue after the device is shut down. |
| Timing and recurring tasksRun tasks by date, time zone, or recurring schedule; only calendar reminders need to be noted. | Documented | Officially supports reminders and recurring checks, and can manage plans. |
| event triggered executionIt is started by events such as emails, messages, Webhooks, and file changes; scheduled polling is a separate… | Unknown | Not researched yet |
| Proactive discovery and follow-upOpportunities, reminders or suggestions are found when no new instructions are received; independent writing requires… | Documented | The official instructions will actively read the connected information in the background and suggest the next step. |
| Memory and personalization | ||
| Long-term memory and preferencesPreserve and use preferences, context, or project state across sessions; don't treat context windows as long-term… | Documented | Official description receives ChatGPT memory and saves long-term working context. |
| Cross-channel continuationDifferent entries can continue the same task and related context; message mirroring and context can be separated. | Documented | The official explanation is that the same dot can use relevant context across channels, and the messages will not all be mirrored. |
| Memory review and correctionUsers can view, correct or selectively forget; distinguish between full account reset and selective deletion. | Unknown | Not researched yet |
| Achievements and realistic tasks | ||
| Research and traceabilitySearch, synthesize and present verifiable sources; source accuracy documented by external review. | Unknown | Not researched yet |
| Documentation and multimodal resultsCreate or modify available products such as documents, tables, presentations, images/audio and videos; format as facet. | Documented | Official instructions can prepare presentations, content and other documentation results. |
| Coding and runnable deliveryImplement, run, test code, and deliver the application or change; only generating snippets is distinguished from… | Documented | Official descriptions enable changes to be implemented and tested, and PRs prepared for user review. |
| Shopping reservation and service processingActual execution of shopping, reservation, unsubscription or service process; stopping at draft, shopping cart,… | Unknown | Not researched yet |
| Collaboration and task closed loop | ||
| Cross-application multi-step processThe same task transfers information and executes across multiple applications; multiple independent plug-ins are not… | Unknown | Not researched yet |
| Parallel tasks and agent collaborationMultiple tasks/agents advance and hand off simultaneously; logging whether files, sessions, and identities are shared. | Documented | The official description states that background tasks can be advanced in parallel while the user continues the conversation. |
| Failure recovery and waiting for follow-upIt can be recovered after failure, waiting for reply or external changes; the completed effect must be supported by… | Unknown | Not researched yet |
| Authorization, security and auditability | ||
| Fine-grained authorizationAllow, deny, or request confirmation by action, object, and scope; presence of controls does not equal security score. | Documented | Officially provides custom action boundaries and approval checks. |
| Manual takeover and continuationIt can be handed over to the user to complete the sensitive/blocked steps, and then returned to the agent to continue. | Documented | The official provides cloud computer takeover and return control. |
| Activity and result trackingDisplay tool actions, execution records, receipts, task status or sources; the display range is noted separately. | Documented | Official portal to mission activities, results, and pending requests. |
| Export, Undo and DeleteSupports data export, connection cancellation or task/data deletion; the three types of operations retain evidence… | Unknown | Not researched yet |
Similar agents
All OpenAI dots alternatives →