{"collection":{"slug":"frontier-ai-labs","name":"Frontier AI Labs"},"digest":{"id":"cwd_uBNadxbw_D8e4OFUpPKrH","weekStart":"2026-09-07","title":"GPT-6 Astra lands everywhere, and voice stops being a side channel","intro":"OpenAI shipped GPT-6 Astra as its most capable business model and made GPT-Live 1 generally available for full-duplex voice, while launching ChatGPT for Financial Services and a managed Agents API. Anthropic's Claude Code picked up a plugin eval suite and a sandbox containment fix.","body":"OpenAI had one of its heaviest weeks of the year, and the releases rhyme: the company is collapsing the distance between what a model can do in text and what it can do in a live conversation, an agent loop, or a vertical product.\n\n### Astra arrives, and voice becomes a first-class surface\n\nThe centerpiece is [GPT-6 Astra](/release/rel_rdfIKNBfwx9_UMZQ5szGS-openai-launches-gpt-6-astra-its-most-capable-model-for-business), billed as OpenAI's most capable model for business, with stronger reasoning, computer use, and — notably — better writing and design judgment. That last pair is a tell: this is a model pitched at deliverables, not benchmarks.\n\nIt shows up quickly in voice. [ChatGPT Voice now runs GPT-5.6 or GPT-6 Astra](/release/rel_drmj0_Ve4Le_cbUx1u2q7-chatgpt-voice-adds-gpt-5-6-and-gpt-6-astra-simplifies-gpt-live-daily-limits) for search and harder reasoning, with model and reasoning effort chosen through the same controls as text chat; the separate Instant/Medium/High Voice intelligence levels are deprecated. Enterprise and Edu workspaces got the same treatment alongside new Live pricing, where credit-based billing drops to [1 credit per minute from 5](/release/rel_2N9AfjCkDkonhiCowjKMb-chatgpt-enterprise-edu-voice-now-offers-gpt-5-6-gpt-6-astra-new-live-pricing-and).\n\nOn the API side, [GPT-Live 1 is generally available for full-duplex voice](/release/rel_Nf24bFTlBOvkh66kblY3U-openai-api-makes-gpt-live-1-generally-available-for-full-duplex-voice), letting conversations continue while a backend model or agent handles reasoning and tools — via Responses delegation or delegation to your own backend. Voice sessions run $0.05 per minute billed per second, with backend usage charged separately. A companion [product announcement](/release/rel_BQ8Ipf9yMhVhdj_sHnsve-openai-launches-gpt-live-1-in-the-api-for-full-duplex-voice-conversations) covers custom voices and telephony support. The SDKs caught up immediately: [Python](/release/rel_I41W_FgTD5bXGVOfGacci-openai-python-sdk-v3-12-0-adds-live-api), [Node](/release/rel_wuhemNfdO_2KwlVUGFaH--openai-node-sdk-v7-14-0-adds-live-api), and [Go](/release/rel_tiQ7qeRQsDuQDTBmQQbmo-openai-go-sdk-v3-60-0-adds-live-api) all published Live API support on the same day.\n\n### Agents become a managed service\n\nThe other structural release: the [Agents API entered public beta](/release/rel_J4EduPu-D361yCaoxd3g9-openai-agents-api-enters-public-beta-with-managed-codex-harness), built on a managed Codex harness that handles session orchestration, context compaction, and recovery. Durable sessions carry work across turns with progress streaming, custom tools and MCP servers, and sandboxes either OpenAI-hosted or from your own infrastructure — with a [matching product-side launch](/release/rel_GKMngVz-oxaYuWAzAAHoC-openai-launches-agents-api-for-building-cloud-agents). It's supported in the [Python SDK's v3.13.0](/release/rel_uQq80YUnsKaoGI7Y-f7JL-openai-python-sdk-v3-13-0-adds-agents-api) and the [Go SDK's v3.61.0](/release/rel_YVTQkFVTyJk59I1ETgCVg-openai-go-sdk-v3-61-0-adds-agents-api).\n\nAgents need guarded credentials, and OpenAI answered that too: [project API keys now support configurable expiration dates](/release/rel_qiAoNlEJzT5H_6IEhAQkI-openai-api-keys-get-configurable-expiration-dates-and-org-level-max-lifetime), with org- and project-level maximum lifetimes admins can enforce. The [Node SDK](/release/rel_CYxAWaPKmfUyE0BHP6rV4-openai-node-sdk-v7-13-0-adds-api-key-expiration-controls) and [Go SDK](/release/rel_vFKOgXN0yePCYK8oFCAoc-openai-go-sdk-v3-59-0-adds-project-api-key-expiration-support) followed with the corresponding controls.\n\n### Verticals, workspace plumbing, and Claude Code's cleanup\n\n[ChatGPT for Financial Services](/release/rel_DSV5r57edFaWwq3AbM5nz-openai-launches-chatgpt-for-financial-services-with-gpt-6-astra) pairs GPT-6 Astra with built-in financial data for research, modeling, and client-ready materials — the clearest signal yet that OpenAI is packaging models as industry products, not just endpoints. Alongside it came [ChatGPT Images 2.5](/release/rel__gi4cohjeV_d2obkWYpuO-chatgpt-images-2-5-creates-more-polished-personalized-images) and the API-side [GPT Image 2.5 Sunburst and Flare](/release/rel_O8tPqbcLroBS1QCl1YlQP-openai-api-launches-gpt-image-2-5-sunburst-and-flare), with new xhigh and max quality settings.\n\nMost of the remaining churn is workspace-facing: the [Data plugin](/release/rel_HB0ipyauA4km_RkSIrCDa-chatgpt-data-plugin-analyzes-connected-business-data-in-work-and-codex) for analyzing connected business data, [Deep Research coming to Work and Codex](/release/rel_xq5JG61lAEb6zdMsbA38v-deep-research-launches-in-chatgpt-work-and-codex), [Box, Dropbox, and SharePoint in ChatGPT Library](/release/rel_R2J2B8H95u5lKWBjJ56ho-chatgpt-library-adds-box-dropbox-and-sharepoint-integrations), plus group managers and Groups Admin API automation. The bulk of it was duplicated across Business, Enterprise, and Edu editions.\n\nAnthropic's week was quieter but substantive. [Claude Code v2.1.269](/release/rel_TGkQfJi7pkUbEYAUfxqGq-claude-code-v2-1-269-adds-plugin-eval-suite-and-output-style-switching) added `claude plugin eval` for scored, reproducible plugin test suites and `/output-style` switching that works over Remote Control and headless sessions. Before it, [v2.1.268](/release/rel_6ZRH2i6CV0d7Up8LiaCd5-claude-code-2-1-268-fixes-http-400-on-third-party-endpoints-and-webfetch-hangs) fixed an HTTP 400 that broke every turn on third-party Anthropic-compatible endpoints since 2.1.265, and [v2.1.267](/release/rel_AfRurTQfcz1JD3wHLf2cL-claude-code-v2-1-267-adds-effort-caps-fixes-sandbox-bypass-and-session-resume) closed a sandbox containment bypass on macOS and Linux while adding a `maxEffortLevel` cap that applies across Bedrock, Vertex, and Foundry.","releaseIds":["rel_rdfIKNBfwx9_UMZQ5szGS","rel_drmj0_Ve4Le_cbUx1u2q7","rel_2N9AfjCkDkonhiCowjKMb","rel_Nf24bFTlBOvkh66kblY3U","rel_BQ8Ipf9yMhVhdj_sHnsve","rel_I41W_FgTD5bXGVOfGacci","rel_wuhemNfdO_2KwlVUGFaH-","rel_tiQ7qeRQsDuQDTBmQQbmo","rel_J4EduPu-D361yCaoxd3g9","rel_GKMngVz-oxaYuWAzAAHoC","rel_uQq80YUnsKaoGI7Y-f7JL","rel_YVTQkFVTyJk59I1ETgCVg","rel_qiAoNlEJzT5H_6IEhAQkI","rel_CYxAWaPKmfUyE0BHP6rV4","rel_vFKOgXN0yePCYK8oFCAoc","rel_DSV5r57edFaWwq3AbM5nz","rel__gi4cohjeV_d2obkWYpuO","rel_O8tPqbcLroBS1QCl1YlQP","rel_HB0ipyauA4km_RkSIrCDa","rel_xq5JG61lAEb6zdMsbA38v","rel_R2J2B8H95u5lKWBjJ56ho","rel_TGkQfJi7pkUbEYAUfxqGq","rel_6ZRH2i6CV0d7Up8LiaCd5","rel_AfRurTQfcz1JD3wHLf2cL"],"releaseCount":70,"generatedAt":"2026-09-14T06:18:16.453Z","releases":[{"id":"rel_rdfIKNBfwx9_UMZQ5szGS","title":"OpenAI launches GPT-6 Astra, its most capable model for business","path":"/release/rel_rdfIKNBfwx9_UMZQ5szGS-gpt-6-astra-most-capable-business-model","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel_drmj0_Ve4Le_cbUx1u2q7","title":"ChatGPT Voice adds GPT-5.6 and GPT-6 Astra, simplifies GPT-Live daily limits","path":"/release/rel_drmj0_Ve4Le_cbUx1u2q7-voice-gains-gpt-5-6-gpt-6-astra-gpt-live-limits-simplified","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel_2N9AfjCkDkonhiCowjKMb","title":"ChatGPT Enterprise & Edu Voice now offers GPT-5.6/GPT-6 Astra, new Live pricing and usage limits","path":"/release/rel_2N9AfjCkDkonhiCowjKMb-voice-adds-gpt-5-6-gpt-6-astra-live-pricing-and-limits-updated","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_Nf24bFTlBOvkh66kblY3U","title":"OpenAI API makes GPT-Live 1 generally available for full-duplex voice","path":"/release/rel_Nf24bFTlBOvkh66kblY3U-gpt-live-1-ga-full-duplex-voice-at-0-05-min","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel_BQ8Ipf9yMhVhdj_sHnsve","title":"OpenAI launches GPT-Live-1 in the API for full-duplex voice conversations","path":"/release/rel_BQ8Ipf9yMhVhdj_sHnsve-gpt-live-1-brings-full-duplex-voice-to-the-api","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel_I41W_FgTD5bXGVOfGacci","title":"OpenAI Python SDK v3.12.0 adds Live API","path":"/release/rel_I41W_FgTD5bXGVOfGacci-live-api-added-asyncstream-gets-aclose","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_wuhemNfdO_2KwlVUGFaH-","title":"OpenAI Node SDK v7.14.0 adds Live API","path":"/release/rel_wuhemNfdO_2KwlVUGFaH--live-api-support-added","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_tiQ7qeRQsDuQDTBmQQbmo","title":"OpenAI Go SDK v3.60.0 adds Live API","path":"/release/rel_tiQ7qeRQsDuQDTBmQQbmo-live-api-support-added","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_J4EduPu-D361yCaoxd3g9","title":"OpenAI Agents API enters public beta with managed Codex harness","path":"/release/rel_J4EduPu-D361yCaoxd3g9-agents-api-public-beta-managed-codex-harness-durable-sessions","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_GKMngVz-oxaYuWAzAAHoC","title":"OpenAI launches Agents API for building cloud agents","path":"/release/rel_GKMngVz-oxaYuWAzAAHoC-agents-api-managed-cloud-agents-via-codex-harness","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_uQq80YUnsKaoGI7Y-f7JL","title":"OpenAI Python SDK v3.13.0 adds Agents API","path":"/release/rel_uQq80YUnsKaoGI7Y-f7JL-agents-api-added-to-openai-python-sdk","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_YVTQkFVTyJk59I1ETgCVg","title":"OpenAI Go SDK v3.61.0 adds Agents API","path":"/release/rel_YVTQkFVTyJk59I1ETgCVg-agents-api-added-to-go-sdk","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_qiAoNlEJzT5H_6IEhAQkI","title":"OpenAI API keys get configurable expiration dates and org-level max lifetime","path":"/release/rel_qiAoNlEJzT5H_6IEhAQkI-api-keys-now-expire-admins-can-cap-key-lifetime","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_CYxAWaPKmfUyE0BHP6rV4","title":"OpenAI Node SDK v7.13.0 adds API key expiration controls","path":"/release/rel_CYxAWaPKmfUyE0BHP6rV4-api-key-expiration-controls-added","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_vFKOgXN0yePCYK8oFCAoc","title":"OpenAI Go SDK v3.59.0 adds project API key expiration support","path":"/release/rel_vFKOgXN0yePCYK8oFCAoc-project-api-keys-now-support-expiration","org":{"slug":"openai","name":"OpenAI"},"importance":3},{"id":"rel_DSV5r57edFaWwq3AbM5nz","title":"OpenAI launches ChatGPT for Financial Services with GPT-6 Astra","path":"/release/rel_DSV5r57edFaWwq3AbM5nz-chatgpt-for-financial-services-launches-with-gpt-6-astra","org":{"slug":"openai","name":"OpenAI"},"importance":5},{"id":"rel__gi4cohjeV_d2obkWYpuO","title":"ChatGPT Images 2.5 creates more polished, personalized images","path":"/release/rel__gi4cohjeV_d2obkWYpuO-images-2-5-delivers-more-personalized-results","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_O8tPqbcLroBS1QCl1YlQP","title":"OpenAI API launches GPT Image 2.5 Sunburst and Flare","path":"/release/rel_O8tPqbcLroBS1QCl1YlQP-gpt-image-2-5-sunburst-and-flare-launch","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_HB0ipyauA4km_RkSIrCDa","title":"ChatGPT Data plugin analyzes connected business data in Work and Codex","path":"/release/rel_HB0ipyauA4km_RkSIrCDa-data-plugin-brings-business-data-analysis-to-chatgpt-work-and-codex","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_xq5JG61lAEb6zdMsbA38v","title":"Deep Research launches in ChatGPT Work and Codex","path":"/release/rel_xq5JG61lAEb6zdMsbA38v-deep-research-arrives-in-work-and-codex","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_R2J2B8H95u5lKWBjJ56ho","title":"ChatGPT Library adds Box, Dropbox, and SharePoint integrations","path":"/release/rel_R2J2B8H95u5lKWBjJ56ho-box-dropbox-sharepoint-files-now-in-chatgpt-library","org":{"slug":"openai","name":"OpenAI"},"importance":4},{"id":"rel_TGkQfJi7pkUbEYAUfxqGq","title":"Claude Code v2.1.269 adds plugin eval suite and /output-style switching","path":"/release/rel_TGkQfJi7pkUbEYAUfxqGq-plugin-eval-suite-powershell-background-commands-survive-exit","org":{"slug":"anthropic","name":"Anthropic"},"importance":4},{"id":"rel_6ZRH2i6CV0d7Up8LiaCd5","title":"Claude Code 2.1.268 fixes HTTP 400 on third-party endpoints and WebFetch hangs","path":"/release/rel_6ZRH2i6CV0d7Up8LiaCd5-third-party-endpoint-400s-fixed-webfetch-gets-300s-deadline","org":{"slug":"anthropic","name":"Anthropic"},"importance":3},{"id":"rel_AfRurTQfcz1JD3wHLf2cL","title":"Claude Code v2.1.267 adds effort caps, fixes sandbox bypass and session resume bugs","path":"/release/rel_AfRurTQfcz1JD3wHLf2cL-effort-caps-added-sandbox-containment-bypass-fixed","org":{"slug":"anthropic","name":"Anthropic"},"importance":3}]}}