OpenAI Release Notes

Go from idea to a working agent faster with the Agents API.

Build and run cloud agents with the Codex harness, fully managed by OpenAI.

We handle orchestration, long-running sessions, and context management. You focus on what makes your agent unique.

Agents API is available in public beta today to all developers. There are no additional fees for using the Agents API – you simply pay for the tokens and tools your agents use, as outlined on our pricing page.

Explore the Agents API overview to learn more, or follow the quickstart to get started and bring the harness behind Codex into your own agents.

OpenAI runs the agent loop on its infrastructure, coordinating model calls, tool use, and context.

You control the agent’s capabilities and choose where it runs code and works with files.

Your sandbox, your call

Bring your own sandbox or connect a sandbox provider.

Choose the environment that fits your workload:

• CPU, GPU, and memory options
• Fully managed environments or deployments within your VPC
• File and secret storage options

We offer first-class integrations with Blaxel AI, Cloudflare Dev, Daytona, DigitalOcean, E2B, Modal, Oracle Cloud, Runloop AI, and Vercel.

We’re also introducing OpenAI-hosted sandboxes.

Your agent can:

• Run code
• Work with files
• Produce artifacts

Supply files, install packages, and add skills and plugins while we provision and manage the sandbox.

Introducing the Agents API

OpenAI-hosted environments

Self-hosted environments

OpenAI Codex on GitHub

Agents API overview

Agents API quickstart

The Data plugin helps you analyze connected business data in ChatGPT Work and Codex. Ask a business question, investigate changes, or create an interactive dashboard or report. You can refine the analysis with follow-up questions and bring in your team’s metric definitions and other business context.

Administrators can select Data in Workspace settings > Plugins to review its installation policy and manage access by role or group. Data-source plugins and their apps may need separate setup. Members can start with @Data; connected-account permissions and workspace access controls apply. Learn more.

The Data plugin helps you analyze connected business data in ChatGPT Work and Codex. Ask a business question, investigate changes, or create an interactive dashboard or report. You can refine the analysis with follow-up questions and bring in your team’s metric definitions and other business context.

Workspace administrators can manage Data’s installation policy in Workspace settings > Plugins and enable the data-source plugins and apps the team needs. Members can start with @Data once Data is available in their workspace. Connected-account permissions and workspace access controls apply. Learn more.

Performance on real tasks

GPT-Live-1 brings ChatGPT’s natural, full-duplex voice conversations to the API. It listens and speaks simultaneously, handles pauses, interruptions, and backchannels, and adapts when a conversation changes direction.

The model manages the live conversation while delegating deeper reasoning and actions to your choice of backend model, tools, or agent framework.

Paired with GPT-6 Astra at medium reasoning effort, GPT-Live-1 completed 83.6% of Tau3 tasks on the first attempt, versus 45.7% for GPT-Realtime-2.1. Tau3 covers airline, retail, and telecom support.

The same pairing scored 38.1% on TauBanking, which tests document retrieval and account-tool use.

Voices and production features

GPT-Live-1 adds 12 real-time voices:
Quartz · Ripple · Vesper · Willow · Stone · Gleam · Meridian · Bossa · Tempo · Beacon · Delta · Cinder

The expanded selection covers more accents, dialects, and languages. System prompts can shape tone, pace, speaking style, and conversational behavior.

Listen to the new voices

For production applications, the model provides native ASR transcripts and response text, keyword biasing, explicit turn detection, and improved handling of silence and background noise.

Connect using WebRTC for browsers, WebSockets for server-side audio, or Telephony and SIP for phone agents.

Delegation and Codex

Use managed Responses delegation or connect an existing model, agent, or service through client delegation. Your application controls permissions and durable task state; interrupting speech does not automatically cancel backend work.

You can also pair GPT-Live-1 with the Codex SDK. Your app passes conversation context to a Codex thread and returns the result to the active voice session, allowing Codex to investigate a repository or complete delegated work while the conversation continues.

See the GPT-Live-1 and Codex integration

Pricing

Voice sessions cost $0.05 per minute, billed per second. Backend model and tool usage is billed separately.

Read the launch post · Get started with GPT-Live

Building with GPT-Live-1? Share what you learn about interruptions, delegation, and production tool use.

Data plugin in ChatGPT Work and Codex

The Data plugin helps you analyze connected business data in ChatGPT Work and Codex. Ask a business question, investigate changes, or create an interactive dashboard or report. You can refine the analysis with follow-up questions and bring in your team’s metric definitions and other business context.

To get started, install Data from the plugin directory, then start a conversation with @Data. Connected sources may require setup or authorization, and queries use your connected account’s existing permissions.

Box, Dropbox, and SharePoint are now in Library

Box, Dropbox, and SharePoint now join Google Drive in ChatGPT Library. Once connected, you can browse and search available files and folders, then add the files you need to a conversation through Add from Library or @mentions—without uploading them again. You can review supported file previews beside the conversation and follow citations back to the source as you work.

When you’re working with a file, you can now keep Box, Dropbox, and Sharepoint files open beside the conversation while asking ChatGPT to summarize, analyze, compare, or create something new from them. You can also select a folder and ask ChatGPT to work across the files it contains. Your content stays connected to their original source, so it’s easy to return to the original.

Availability

Box, Dropbox, and Sharepoint integrations are now rolling out to Go, Plus, Pro, Business, Edu, Healthcare and Enterprise users on the web in both Chat and Work. Existing file permissions and workspace controls apply. Mobile support will follow.

Rolling out to Plus, Pro, Enterprise, Edu, Healthcare and Business users on the web in both the Chat and Work toggles. Mobile support will follow.

Meet the new Data agent in ChatGPT Work: turn your company’s data into answers, interactive dashboards, and action, just by asking.

People across every business have questions that data can answer. Why did sales slow down? Where is spending rising? Which issues threaten renewals in our largest accounts, and what should we fix first? Getting those answers often means waiting for a report or asking someone else to run the analysis.

We’re introducing a new Data agent in ChatGPT Work⁠ (opens in a new window) to help more people answer those questions themselves. It connects to your company data, investigates what changed, and builds interactive dashboards you can share. Direct and refine the analysis in one conversation, without writing queries or learning a new analytics tool.

Connect to the company data and context your business trusts

The Data agent connects to approved data sources including Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, and more. It can also bring files and documents from Google Drive and SharePoint into the analysis.

It uses your organization’s business terms, metric definitions, custom calculations, and data relationships to interpret the data. This context comes from semantic layers and trusted sources such as Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon, and BI dashboards.

Enterprise administrators choose which data connections are available and which roles can use them. Queries enforce the connected account’s existing permissions, including table, row, and column restrictions.

Turn questions into analysis and action

Ask follow-up questions to investigate the results and review the evidence behind each finding.

Turn the analysis into an interactive dashboard with built-in visualizations. Your team can edit, share, and refresh it as needed. Share your brand guidelines to tailor outputs to your organization’s look and feel.

The Data agent can also build and interact with dashboards in Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot. Direct the work in plain language in the tools your team already uses.

Ask ChatGPT Work to recommend next steps and identify who needs to be involved. It can share the findings through Slack or email and carry out the actions you approve through connected tools.

Built from the tools we use to analyze data at OpenAI

We use the capabilities behind the Data agent broadly across OpenAI. Nearly all of our product team and over two-thirds of our GTM organization use data agents in ChatGPT Work to analyze company data themselves. Our data team made this possible by creating shared business definitions, setting access rules, and putting safeguards in place for sensitive data. Learn more by reading this post⁠ (opens in a new window) and attending our webinar⁠ (opens in a new window).

NTT Data, Thermo Fisher, ServicePiston, and other organizations in our Alpha program are using the Data agent in ChatGPT Work to analyze sales and spending, catch reporting errors, and decide which opportunities to pursue and how to staff them.

Get started with the Data agent

You’ll find the Data agent listed as Data (opens in a new window) in the Plugins directory in ChatGPT Work. Administrators can make it available or install it for their teams through Workspace settings > Plugins. They can also enable and configure the relevant data-source plugins, such as Databricks and Snowflake, and manage who can use them.

If Data isn’t already installed, find it in the Plugins directory and select Install plugin, or go directly to the Data listing (opens in a new window). Complete any required account-connection steps, then start a conversation with @Data and ask your business question.

Try these prompts

Diagnose a metric change

@Data Diagnose why weekly active users changed last week. Identify likely drivers, compare against prior periods, and recommend the next checks.

Design KPI framework

@Data Design a KPI framework for this new product area with primary metrics, drivers, guardrails, targets, and data validation needs.

Create leadership readout

@Data Turn this month’s metrics into a leadership-ready update with actuals, comparisons, drivers, caveats, and recommended actions.

ChatGPT for Financial Services

We’re introducing ChatGPT for Financial Services, a tailored ChatGPT Work experience that combines built-in financial data with GPT‑6 Astra’s reasoning to help teams develop research, financial models, and customized client materials.

This product has been shaped by our design partnership with Morgan Stanley and Evercore. The collaboration has enabled us to pinpoint where OpenAI can solve the biggest challenges for financial institutions and has guided us on the solutions that will help every banker in their day-to-day work.

Built-in premium data from providers like Daloopa, PitchBook, LSEG News, and Crunchbase remove the challenges with MCP connectors and access to data. This data is indexed and hosted by OpenAI to enable higher accuracy and new features like granular citations so that bankers can trace figures and claims back to their sources, and check the evidence as their analysis develops.

The product also offers state of the art frontier intelligence, including GPT‑6 Astra, natively and will continue to have newer models available out of the box as they are released. Firms can also centrally manage access and data connections, supported by ChatGPT’s enterprise security and governance controls.

Our early work with Morgan Stanley and Evercore has helped steer where we have started: investment banking and equity research. Reliable access to data and high quality artifact creation proved to be the biggest pain points for their teams. Our work with partners will inform post training, product improvements, and our expansion into other financial services categories.

Design partners

Morgan Stanley: “The promise of frontier research becomes real when it helps our people do the work that matters for our clients. We’re working alongside OpenAI to bring that intelligence into how we research companies, develop analysis, and prepare advice. We can test ideas together and help shape how the technology develops. For our firm, that means a product increasingly informed by our people, our methods, and the work that matters to our clients.”

Evercore: “At Evercore, we focus on the questions that matter most to our clients, and we have an opportunity to apply frontier intelligence to help address them. We are working closely with OpenAI to shape how its technology can deepen the insights behind our advice, while building on the judgment and rigorous standards our clients expect from Evercore.”

ChatGPT for Financial Services brings together the financial data teams need, with the depth of detail expected. We’ve included premium financial data, streamlined existing provider connections, and improved MCP performance.

ChatGPT for Financial Services includes datasets from providers like Daloopa, PitchBook, LSEG News and Crunchbase covering earnings transcripts, financial statements, company fundamentals, private companies, and more.

Teams can start working with these datasets immediately, with no separate contracts to negotiate or connectors to set up. We index and host this data on OpenAI infrastructure, allowing us to improve retrieval, latency, and how we surface it in the product experience.

For a banker running a P&L normalization analysis, that means being able to inspect the reconciliation and notes behind an adjusted EBITDA, understanding which costs were excluded, and deciding how to use it in a valuation.

Our partners are central to this work and as we expand coverage, we will continue to deepen our models’ understanding of these datasets. We will post train our models to find, interpret, and use this data like we know the best analysts can.

Trace figures and claims to specific tables and passages, with the supporting information highlighted for review.

Data providers include Crunchbase, Daloopa, and PitchBook with testimonials highlighting the value and reliability of their data.

We know many customers already have existing data subscriptions. So we’re working with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody’s on shared sign-in and entitlement integrations. Providers will be able to recognize users through their ChatGPT sign-in to enable automatic access to data they’re already entitled to.

Subscription providers include S&P Global and LSEG with statements emphasizing trusted data and collaboration.

It can be challenging to effectively use MCP connectors for data providers. We have focused on making some of the most used MCPs in financial services like S&P Global and FactSet optimized for immediate use in the product, so teams can spend less time troubleshooting requests and more time on analysis.

Our broader connector ecosystem includes over 50+ connectors including Datasite, Box, Preqin, and Intapp.

We optimized the connector reliability for popular MCP connectors through the process of automated evaluation and iteration.

GPT‑6 Astra is state of the art across three of the core capabilities required for financial services work: information retrieval, financial reasoning, and artifact generation.

It can navigate and understand figures, tables, and the supporting notes in financial documents. It can take inputs, run financial analysis, and draw conclusions. And it can take all of this work and accurately synthesize it into documents, spreadsheets, and slides.

Benchmarks show GPT‑6 Astra scoring 69.9% on OfficeQA Pro compared with 60.2% for GPT‑5.6 Sol.

We’ve built fundamental capabilities into the product so that it can research across multiple sources, trace figures across different periods, and interpret annotations in public data. Once that analysis is complete, teams can build interactive charts and visualizations.

Compare company performance in interactive charts, with the underlying data and sources available for review.

In ChatGPT for Financial Services, administrators can publish Excel, Word, and PowerPoint templates through a dedicated admin page. With firm templates and style guides configured, teams can turn their analysis into valuation models, research notes, and pitchbooks in their firm’s format and style.

Publish firm templates to the teams who use them.

Protecting material non-public information and client confidentiality is critical for financial institutions. ChatGPT for Financial Services builds on ChatGPT Enterprise’s SAML SSO, SCIM provisioning, and role-based access controls.

Your firm’s business data is not used to train our models by default. It is encrypted at rest and in transit, and admins can configure workspace retention. Compliance teams can export supported workspace logs through the OpenAI Compliance Platform into their existing audit and investigation workflows.

Firms can manage access to skills and apps by role and enable or disable supported app read and write actions. Multiple workspaces can also be created to enforce information barriers.

ChatGPT for Financial Services is one way we serve customers across the industry, but we recognize that it will require a range of solutions to address the needs of the finance industry. OpenAI has a long history of collaborating with innovators to unlock novel AI solutions. Financial services firms and developers can use our API to build specialized applications for the needs they understand best. OpenAI brings frontier models and the capabilities to put them to work. Financial institutions, data providers, and software partners bring specialized expertise, trusted information, and customer relationships.

ChatGPT for Financial Services is available to eligible financial institutions. If you’re interested, please contact us or reach out to your account team.

Equip your teams with GPT-6 Astra, included financial data, and new tools for research, financial modeling, and client materials.

GPT‑Live‑1 brings ChatGPT’s natural, full-duplex conversations to the API, with more control over how voice agents speak and act.

We’re launching GPT‑Live‑1 in the API, giving developers a powerful, natural voice model for building voice-enabled apps and business workflows. First introduced in ChatGPT, GPT‑Live‑1 is capable of listening and speaking at the same time, and, as seen with Codex and ChatGPT Work (opens in a new window), can delegate deeper reasoning and actions to the models and tools it is paired with.

For the API release of GPT‑Live‑1, we’ve focused on new capabilities that let developers steer and customize voice experiences around their users, workflows, and goals. A core GPT‑Live‑1 strength, smooth interruption handling, is already delivering business impact: in early evaluations, Speak found that GPT‑Live‑1 gave learners more time to think before the language tutor responded, cutting interruptions by almost 80% versus previous turn-based systems.

Key strengths of GPT‑Live‑1 in the API:

• Interruption handling: Improves interruption handling via a single model that reasons over incoming and outgoing audio together, avoiding the latency and brittle handoffs of chained STT–LLM–TTS architectures.
• Reasoning & tool calling delegation: GPT‑Live‑1 can delegate reasoning and tool calls to a backend text model like GPT‑6 Astra or a third-party model.
• Tone, pace, and style: Lets developers shape an agent’s tone, pace, and conversational style through the system prompt.
• Silent context management & background noise: Better handles background noise and silence without interrupting the conversation or narrating every step out loud.
• Long-session reliability: Improves context retention and conversational quality across extended interactions.
• Telephony support: Enables deployment of full-duplex voice agents for phone calls, from restaurant reservations to customer support.

Simplify your voice-agent architecture and reduce voice latency

Traditional voice agents stitch together speech-to-text, a reasoning model, and text-to-speech. Each handoff adds latency and creates more opportunities to lose timing, context, or the natural rhythm of a conversation. Developers are often the ones left coordinating those stages, including what happens when someone interrupts, pauses, or changes direction.

GPT‑Live‑1 handles listening and speaking in a single model, simplifying the voice layer. It can respond to interruptions and acknowledgements as they happen, while delegating deeper reasoning to the back end. This lets the conversation continue while work happens in the background.

“Compared to our cascaded build, GPT‑Live‑1 simplified our code base by 80% and removed 23K lines of code. This enabled natural, real-time patient conversations & freed our team to improve the experience from booking an appointment to navigating care.”
—Tony Stoyanov, Co-Founder & CTO

Developers choose the models, tools, and agent harness behind the conversation. For example, they might pair GPT‑Live‑1 with a model like Luna for high-volume tasks like scheduling or order updates, and use a model like Astra for complex customer issues that require reasoning. That flexibility lets developers match reasoning depth, speed, and cost to each task.

GPT‑Live‑1 natively provides ASR transcripts and response text. It also offers strong alphanumeric understanding and supports keyword biasing. Although GPT‑Live‑1 is not a turn-based model, it natively supports turn detection, so developers can continue to build around explicit turn boundaries.

Measuring the full-duplex advantage

Across our evaluations, GPT‑Live‑1 improves Full Duplex Bench performance by 30 percentage points over GPT‑Realtime‑2.1, with large gains in turn-taking latency and interactive behavior. Paired with GPT‑6 Astra at medium reasoning effort, it also ranks #1 on Tau3, which measures frontier voice-agent intelligence on end-to-end tasks.

What customers are saying

“Adding GPT‑Live‑1 into Yelp Host and Hatch improved turn-taking and accuracy over our traditional voice architecture. When Yelp Host uses GPT‑Live‑1 to answer calls, like reservations and food orders, we're seeing meaningful improvements in call handling rates. Callers are also speaking fuller, more natural sentences, which tells us the experience on the other end of the phone feels genuinely different.”
—Alex Levy, Chief Technology Officer

New voice options

Developers need voices that fit their product and sound natural to the people using it. With GPT‑Live‑1, we’re expanding from a small set of real-time voices to a broader selection across accents, dialects, and languages giving developers more choice in how their assistants sound.

We’ll continue to expand voice options and language availability over the coming months.

Pricing & Availability

GPT‑Live‑1 is available in the API today⁠ (opens in a new window) at $0.05 per minute for the front-end voice layer. Pair it with the backend model and agent harness that fit your product, then build a voice experience that can scale with the work it needs to do.

For custom voice access, contact sales to learn more about eligibility and the request process.

Another way to build voice workflows on top of GPT‑Live‑1 is with OpenAI Presence, which uses the model to power real-time voice interactions. Presence helps enterprises deploy trusted AI agents that can answer questions, resolve issues, use company systems, take approved actions, and escalate to people when needed. Reach out to your OpenAI account director to learn more.

As we’ve scaled Codex and ChatGPT for Work to millions of people around the world, we’ve learned what it takes to make long-running agents work well in practice. Useful agents need a powerful harness that manages context, uses tools efficiently, and coordinates subagents. They also need infrastructure that keeps them running reliably for days, with environments where they can work with files, run code, and save intermediate results.

Today, we’re introducing the Agents API (opens in a new window) in public beta, bringing that same harness and infrastructure that powers Codex to developers through a simple, flexible API.

OpenAI hosts and maintains the harness. You choose the agent’s compute environment: in an OpenAI-managed sandbox, on your own infrastructure, or with one of our sandbox partners. The Agents API gives you a strong foundation for building agents on top of our optimized agent harness and infrastructure, so you can focus on the tools, knowledge, and workflows that make your agent unique.

“With the Agents API, our evaluation score went from 0.71 to 0.85. The subagent support in the API is great and drastically sped up our workflow. Previously it was pretty cumbersome to observe and orchestrate subagents in our old setup but the new APIs gave us a 4x latency reduction. We spent a long time trying to optimize for this and the subagent flows were a huge out-of-the-box lift.”

Jack Weissenberger, CTO, Ciridae

Agents API powers your agents with the same harness and infrastructure behind Codex.

Build cloud agents with a single API call

With the Agents API, you can create a production-ready agent in a single API call by specifying the task, model, tools, and environment.

What our customers are saying about Agents API

“Transforming real-world businesses means deploying AI into workflows of every shape. Agents API supplies the harness; the environment, context, and UX stay ours. With our AI platform Nexus we now stand up agents in hours across industries, from residential services to architecture.”

Rasmus Wissmann, CTO, Long Lake

Choose your agent environment

Different workloads need different compute, storage, and deployment options. The Agents API lets you choose a sandbox that fits your application.

We’re partnering with ecosystem providers (opens in a new window), including Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel, to provide first-class integrations for a range of needs:

• Fully managed environments or deployments within your VPC
• Specific file and secret storage mechanisms
• Different CPU, GPU, and memory configurations, with performance, cold-start, and cost profiles to match your company’s workflow.

The Agents API offers first-class integrations with popular ecosystem providers.

OpenAI hosted sandboxes

For developers who want to get started quickly and scale efficiently, we’re also introducing the OpenAI hosted sandbox (opens in a new window). This leverages the same sandboxing infrastructure that powers Codex and ChatGPT.

OpenAI provisions and manages the sandbox, giving your agent a secure and performant environment to run code, work with files, and produce artifacts. These sandboxes can be flexibly configured with your files, packages, skills and plugins to give the agent what it needs to complete the task.

Build with an evolving Codex harness

Taking advantage of new model capabilities often means reworking your harness, taking valuable time away from improving your application. The Agents API provides versioned access to these capabilities with each model launch. We maintain and continuously improve the harness alongside our models, helping your agents get better performance from every upgrade. For example, recent improvements to the harness include:

Keep agents working across long sessions

To support models working for hours, we’ve built context management that helps agents carry relevant information across longer sessions. The Agents API automatically compacts (opens in a new window) earlier context as a session approaches its context limit, preserving information the agent needs to continue. Developers can build workflows that span multiple context windows without implementing their own compaction logic.

Help agents efficiently use more tools

The Agents API helps agents find the right tools and use them efficiently. Tool search (opens in a new window) loads relevant tool definitions as needed, helping reduce token usage and cost while preserving the model’s cache. Once tools are available, programmatic tool calling (opens in a new window) lets agents run calls in parallel, chain related operations, and filter or combine results in code so they can work through large volumes of data while bringing only the relevant results back into context. The Agents API supports MCP, custom functions, and built-in tools like web search.

Let agents parallelize work with subagents

With multi-agent support (opens in a new window), the Agents API can break complex tasks into independent pieces and delegate them to subagents that work in parallel. Each subagent maintains its own context, helping it stay focused on its assignment, while the main agent coordinates their work and brings the results together. This can speed up research, analysis, and coding tasks that benefit from parallel work, without requiring you to build your own orchestration.

Build on an open-source foundation

The Agents API is powered by the open-source Codex harness, giving developers visibility into the core logic that coordinates model calls, tools, and context. With the Agents API, OpenAI operates and maintains that harness while developers can inspect and learn from its public codebase (opens in a new window).

Start building

Agents API is available in public beta today to all developers. There are no additional fees for using the Agents API – you simply pay for the tokens and tools your agents use, as outlined on our pricing page (opens in a new window).

Explore the Agents API overview (opens in a new window) to learn more, or follow the quickstart (opens in a new window) to get started and bring the harness behind Codex into your own agents.

During the public beta, we’ll iterate quickly based on your feedback as we work toward general availability. Let us know what’s working, where you’re running into friction, and what you need to build and run your agents in production.

Install with pip install --upgrade openai-codex==0.154.0 (Python 3.10 or later). This release includes the matching openai-codex-cli-bin==0.154.0 runtime.

• Add max and ultra reasoning-effort values.
• Add ExternalMessage to synchronous and asynchronous run() and turn() calls. External content can start a turn or join an active regular turn with tool-level authority; it does not grant user authorization. Consumers receive independent event streams.
• Add include_turns on resume/fork, turn_service_tier for one newly started turn, and source metadata. History selection changes the returned response, not model context. Existing defaults are preserved when these options are omitted.
• Refresh generated protocol models and notifications, and preserve completion events that arrive before a turn-start response.

Check these migrations when upgrading:

• HookMetadata wraps its handler in .root. Replace accesses such as hook.command with hook.root.command, checking hook.root.handler_type before reading handler-specific fields.
• Some previously unknown notifications now have typed payloads. Read named fields instead of .params; unknown or invalid payloads still use UnknownNotification.
• Manually constructed or late-joining turn handles receive events from their attachment point. Earlier output is not replayed, so collected results can be partial; attaching after completion can raise TransportClosedError. Use thread.read(include_turns=True) for saved history. Handles returned directly by thread.turn(...) retain events from when their request is sent.
• Custom codex_bin overrides require CLI 0.151.0 or newer for ExternalMessage and the new history/per-turn options.

Changelog

Full Changelog: rust-v0.153.0...rust-v0.154.0

• Fix punctuation in npm packaging documentation
• Use native spawning for bare macOS MCP commands
• Add Vim replace mode to the TUI composer
• Add managed worktree creation
• Refactor shared TUI input routing
• Separate TUI preferences from server configuration
• Add macOS voice runtime projection
• Retry TUI reconnects while threads are closing
• Add GNU Linux voice runtime preparation
• Add Windows voice runtime preparation
• Show live context compaction status in the TUI
• Support the app-server daemon on Windows
• Honor explicit plugin mentions during MCP startup
• Harden embedded composer input handling
• Allow reviewing and continuing misalignment-paused chats
• Existing sessions pick up newly installed plugin tools and refresh skills and hooks after external plugin upgrades or rollbacks
• MCP connections coordinate OAuth token refreshes and surface login challenges when refresh fails, without automatically replaying rejected tool calls
• Startup avoids running workspace-controlled helpers before trust is established, and the macOS sandbox blocks terminal input injection
• Remote resume and fork operations preserve saved permissions; fresh sessions and forks respect server model defaults unless explicitly overridden
• Resuming a conversation open in another app now shows a read-only transcript with a retry option while preserving your draft
• Automatic approval reviews better preserve authorization context through compaction and reject approvals invalidated by new user instructions or answers
• Updated the bundled OpenAI Docs skill with GPT-6-Astra migration, compatibility, and prompting guidance
• The deprecated codex mcp-server entry point is no longer available.

GPT‑6 Astra: The next generation in intelligence for work

Our most capable model, built for all the work businesses need to get done.

Last week we introduced GPT‑6 Astra, the world’s most intelligent and aligned model, now available in ChatGPT Work, Codex, and the API. Astra is state-of-the-art on computer use, browsing, professional work, software engineering, cybersecurity, and science, so teams can take on the most demanding professional work with unmatched speed, accuracy, and judgment.

The world’s best model for complex work

Most AI systems require businesses to prepare their data, redesign workflows, and build custom integrations before they can deliver value. Astra changes that. In ChatGPT Work and Codex, it can write code and work through the same applications people use every day—even when those applications don’t have an API. That means businesses can put AI to work within their existing workflows from day one, without extensive preparation or engineering work.

Excel competition:

GPT‑6 Astra can complete Financial Modeling World Cup challenges using computer use about four times as fast as the winning human competitor—helping analysts spend less time building models and more time interpreting results and making decisions. From the 2023 Microsoft Excel World Championship (opens in a new window).

Within the first few days of rollout, we’re already seeing customers put Astra to work, from optimizing GPUs to spotting discrepancies in financial statements to producing more on-brand decks.

“We’re integrating GPT‑6 Astra into Devin’s harness on launch day, where it delivers state-of-the-art performance on our internal testing benchmark. Its excellent computer use, writing, and codebase understanding improved testing right out of the box: videos are noticeably easier to follow, and reports are clearer and more concise”
— Silas Alberti, SVP Research, Cognition

At OpenAI, Astra was rolled out internally weeks before launch, so we saw first hand how bleeding-edge capabilities like computer use could change the way we work. Our developer and marketing teams used Astra and Codex to turn three hours of multicamera footage into our GPT‑6 Astra Developer First Impressions video (opens in a new window) which has already garnered over 550k views in just 4 days. Our engineering team used Astra to uncover and resolve a memory-allocation bottleneck that was causing slow Codex sessions in a test environment. By switching allocators, they were able to produce 25× lower turn latency with roughly 30% higher peak memory use.

Astra is also better at following a company’s voice, templates, and design standards, so the first result is closer to something a team can put to use.

GPT‑6 Astra creates a slideshow about GPT‑Gaia, a fictional model, using just a few slides from OpenAI’s presentation template, capturing the correct tone and layout throughout. This means you can expect slide decks that are correctly formatted for your business standards.

More useful work for every dollar

Astra continues our commitment to providing extremely efficient models that deliver more useful work per dollar to our customers. It's been trained to complete tasks in fewer tokens with fewer retries, which means less rework and lower cost per task. With Astra, OpenAI occupies the majority of the cost-efficiency frontier on professional work and coding evaluations, including Terminal Bench 4.0 and Artificial Analysis Intelligence Index. Pricing starts at $10 per million input tokens and $50 per million output tokens.

Terminal-Bench 4.0 tests agents on complex terminal-based tasks, including software engineering, system configuration, and data analysis. GPT‑6 Astra reaches a new high at 57.9%, compared with 37.3% for GPT‑5.6 Sol and 55.8% for Claude Fable 5.1, at approximately 9% and 63% lower estimated API cost per task, respectively.

“Astra sets a new record on DeepSWE v1.1 at 74%. It did so with fewer steps and greater token efficiency than has ever been achieved by frontier models, especially on complex, long horizon tasks. Certainly, this model will have a noticeable impact on high quality, real-world software engineering.”
— Serena Ge, Co-Founder & CEO, Datacurve

More safety and control for consequential work

Giving AI access to business systems requires confidence in how it will act. With this in mind, Astra is our most aligned model yet, with stronger adherence to human intent and authorization.

During training, we tested Astra on our internal computer use safety benchmark which tests models against the hardest business scenarios such as exposing confidential information, sharing a dashboard too broadly, or deleting data. In this evaluation, Astra produced unintended outcomes 89% less often than GPT‑5.6 Sol and 74.7% less often than Claude Fable 5.1. Additional confirmation and automated review further improved performance for GPT‑6 Astra and GPT‑5.6 Sol.

Organizations can also decide how broadly to deploy Astra. New enterprise admin controls let them restrict access to approved websites and desktop applications, manage uploads and downloads, and control browsing history. ChatGPT Work and Codex also include safeguards such as confirmation policies, which can require approval before consequential actions, and automated review of potentially unsafe or unauthorized tool calls. These controls allow teams to start with a limited configuration and expand access over time.

To further access, alongside Astra, we’re also launching new enterprise plugins in ChatGPT Desktop. Powered by the latest browser use capabilities, plugins from Oracle Analytics, Power BI (a Microsoft Fabric service), Navan, and Avalara make it easier to access familiar enterprise applications.

Astra (opens in a new window) is also the first model to reach the Critical cybersecurity capability threshold under our Preparedness Framework. With that increased capability, we’ve strengthened protections (opens in a new window) against both misuse and the model taking unauthorized actions including training Astra to respect safety and security boundaries, improving its resistance to attempts to bypass safeguards, and deploying automated checks designed to block harmful responses.

Zero Data Retention is available for eligible API customers on supported endpoints, subject to approval.

Start using Astra today

Try GPT‑6 Astra in ChatGPT Work (opens in a new window) or Codex (opens in a new window), or build it into your own products and workflows through the API (opens in a new window).

Enterprise administrators can enable Astra under their applicable rate card and agreement. Enterprise access is off by default at launch.

Try GPT‑6 Astra

ChatGPT Work or Codex (opens in a new window)

API (opens in a new window)

Updated models and usage limits in ChatGPT Voice

ChatGPT Voice can now use GPT-5.6 or GPT-6 Astra when it needs to search or reason through harder questions. Choose your model and reasoning effort using the same controls as text chat. Available models and their usage limits depend on your plan and workspace settings.

We’re also updating GPT-Live pricing and daily usage limits:

• Enterprise workspaces with usage-based billing in U.S. dollars: $0.05 per minute.
• Enterprise and Edu workspaces using credits: 1 credit per minute, down from 5 credits.
• Legacy Enterprise and Edu plans: Up to 3 hours with GPT-Live-1, replacing the previous Live and mini allowances.

Legacy Enterprise and Edu plans no longer switch to GPT-Live mini after reaching a Voice limit. Search and reasoning usage follows the selected model’s normal limits and pricing, separately from Voice usage.

With these updates, the separate Instant/Medium/High Voice intelligence levels are deprecated.

Learn more:

https://help.openai.com/en/articles/20001274-chatgpt-voice

Updated models and usage limits in ChatGPT Voice

ChatGPT Voice can now use GPT-5.6 or GPT-6 Astra when it needs to search or reason through harder questions. Choose your model and reasoning effort using the same controls as text chat. Available models and their usage limits depend on your plan and workspace settings.

We’re also simplifying GPT-Live daily usage limits:

• Business Standard seats: Up to 3 hours with GPT-Live-1.
• Business Premium seats: Up to 15 hours with GPT-Live-1.

Additional Voice usage costs 1 credit per minute, down from 5 credits. Search and reasoning usage follows the selected model’s normal limits and pricing, separately from Voice usage.

With these updates, the separate Instant/Medium/High Voice intelligence levels are deprecated.

Learn more:
https://help.openai.com/en/articles/20001274-chatgpt-voice

Updated models and usage limits in ChatGPT Voice

ChatGPT Voice can now use GPT-5.6 or GPT-6 Astra when it needs to search or reason through harder questions. Choose your model and reasoning effort using the same controls as text chat. Available models and their usage limits depend on your plan.

We’re also simplifying GPT-Live daily usage limits:

• Go: Up to 3 hours with GPT-Live-1 mini, replacing GPT-Live-1 access.
• Plus: Up to 3 hours with GPT-Live-1.
• Pro ($100/month): Up to 15 hours with GPT-Live-1.
• Pro ($200/month): Unlimited GPT-Live-1 usage.

Plus and Pro no longer switch to GPT-Live mini after reaching a Voice limit.

With these updates, the separate Instant/Medium/High Voice intelligence levels are deprecated.

Learn more: https://help.openai.com/en/articles/20001274-chatgpt-voice

Share files and folders from your Library

Library sharing lets you share files and folders in ChatGPT, choose who can access them, and work with shared content directly in your conversations.

What’s new

• Share with specific people.
  Invite recipients and assign Viewer or Editor access.
• Share across your workspace.
  Make content available to everyone in your workspace.
• Manage access.
  Review who has access, change permissions, or remove access from the sharing dialog.
• Find and use shared content.
  Access items in Shared with me and use shared files in conversations with ChatGPT.

Shared folders and ownership

Files uploaded to a shared folder belong to the folder’s owner. The person who uploaded each file is still recorded.

If you upload a file to someone else’s shared folder and they later remove your folder access, the file stays in their folder and you lose access to it.

Deep Research in ChatGPT Work and Codex

Deep research is now available in ChatGPT Work and Codex. Research complex questions across the web, your files, and supported connected apps, then turn the findings into an editable document with citations. You can steer the research as it runs and request a document, presentation, spreadsheet, or Site, depending on available tools.

To start, type @Deep Research in Work or explicitly ask for deep research. Available to Plus, Pro, Business, Enterprise, and Edu users with Work access on web, desktop, iOS, and Android. It uses your existing Work/Codex allowance or credits; deep research limits in Chat are unchanged.

Learn more.

ChatGPT Images 2.5: new ways to create and edit

ChatGPT Images 2.5 brings sharper details, more precise editing, and faster image generation. We’re also introducing new ways to create and share:

• Start from a template.
  Open Images → Templates, choose a template, and customize the details. ChatGPT may ask follow-up questions to help refine your request.

• Turn a sketch into an image.
  On mobile, type
  @
  in the message box and select
  Sketch
  . Draw your idea, then describe the image you want to create.

• Edit and comment.
  On mobile, open a generated image full-screen to edit it or add comments to tell ChatGPT what you’d like to change..

• Share a prompt.
  Share the prompt for a generated image so others can create their own version.

Some users will also see animated dots and a Snake game while images are generated. The loading experience varies by platform and rollout; Snake isn’t available in Work mode.

Templates are not yet available in Work mode. Existing image-generation limits are unchanged.

Learn more about Images in ChatGPT.