---
collection: frontier-ai-labs
collection_name: Frontier AI Labs
week_start: 2026-10-19
title: OpenAI and Anthropic both bet on cheaper, longer-lived context
release_count: 47
generated: 2026-10-26
canonical: https://releases.sh/collections/frontier-ai-labs/digest/2026-10-19
---

# OpenAI and Anthropic both bet on cheaper, longer-lived context

OpenAI made the Decisions API generally available on gpt-6-luna and added 12-hour prompt caching, while Anthropic shipped a breaking API version with major SDK releases and donated Agent Skills to the Agentic AI Foundation.

### Cheaper context, longer cache

Both labs spent the week attacking the cost of keeping context around. OpenAI's API now offers [12-hour prompt caching](https://developers.openai.com/api/docs/changelog#2026-10-20), so shared instructions, tools, and context stay warm even when requests are hours apart — cache writes at that duration cost 2.5x the input token price, a trade worth running the numbers on. The [Decisions API went generally available on gpt-6-luna](https://developers.openai.com/api/docs/changelog#2026-10-22) and took direct advantage of it: shared input can be cached across prompt-cache calls, question instructions can carry images, structured data flows in through `input_json`, and a single request now handles up to 400 questions instead of 200. OpenAI framed the offering simply as [fast, focused, low-cost decisions for any application](https://openai.com/index/decisions-api).

The SDKs were reworked around those Decisions types, and the changes are breaking in minor releases. The [Go SDK moved Decision and Live instructions to typed unions](https://github.com/openai/openai-go/releases/tag/v3.86.0), so plain string assignments stop compiling. The [Node SDK made question instructions a string or parts array](https://github.com/openai/openai-node/releases/tag/v7.43.0), which means `question.instructions.trim()` can now blow up on the array form, and code extending `DecisionInputMessage` has to switch to the namespaced variant. Python took the [same structured instructions and richer input message roles](https://github.com/openai/openai-python/releases/tag/v3.37.0), with string handling preserved but now needing an `isinstance` check. Vertex AI support for the Responses API landed across [Node](https://github.com/openai/openai-node/releases/tag/v7.41.0), [Python](https://github.com/openai/openai-python/releases/tag/v3.36.0), and [Go](https://github.com/openai/openai-go/releases/tag/v3.85.0) — and separately as a marked [Go Vertex client milestone](https://github.com/openai/openai-go/releases/tag/v3/vertex/v1.0.0) — alongside voice consent management, voice deletion, and 30-minute cache-write usage counts. Long tail: [Node's WebSocket truncation fixes](https://github.com/openai/openai-node/releases/tag/v7.40.0), [Go's X.509 and WebSocket restore fixes](https://github.com/openai/openai-go/releases/tag/v3.84.1), [voice consent in Go](https://github.com/openai/openai-go/releases/tag/v3.84.0), [raw WebSocket helpers](https://github.com/openai/openai-go/releases/tag/v3.83.0), and [per-attempt auth timeouts](https://github.com/openai/openai-go/releases/tag/v3.85.1).

### Claude Code's security week

Claude Code shipped a fistful of updates, and the throughline was sandbox and permission hardening. Sandboxed Bash commands on macOS and Linux no longer [run shell startup files outside the sandbox](https://code.claude.com/docs/en/changelog#2-1-309), symlinked `denyWrite` globs now apply at real paths, and sandboxed commands can't touch the MCP config passed via `--mcp-config`. Linux users got [nested user namespaces](https://code.claude.com/docs/en/changelog#2-1-305) as an opt-in — with the honest caveat that it lets them get around the sandbox's file protections — and a fix for a regression where [Vertex and Bedrock requests failed on tool search rejection](https://code.claude.com/docs/en/changelog#2-1-305). Elsewhere: [worktree removal no longer drops commits](https://code.claude.com/docs/en/changelog#2-1-304) when HEAD is detached, plus `/auto-mode-setup`; [a new OpenTelemetry safety event](https://code.claude.com/docs/en/changelog#2-1-307) and `/claude-api api-version` for API version 2026-10-21; [cloud session login and prompt-too-long fixes](https://code.claude.com/docs/en/changelog#2-1-306); [background cleanup patches](https://code.claude.com/docs/en/changelog#2-1-310) plus a Monitor tool on Bedrock, Vertex, and Foundry; and [a round of reliability fixes](https://code.claude.com/docs/en/changelog#2-1-308).

### Anthropic's breaking API version, and an open standard

Anthropic released [API version 2026-10-21 with a migration guide](https://platform.claude.com/docs/en/release-notes/overview#2026-10-21), and every official client now defaults to it: Python and Ruby hit 2.0.0, TypeScript and PHP 1.0.0, Go 2.0.0, Java 3.0.0, C# 13.0.0, and the `ant` CLI 2.0.0. The SDKs carry breaking changes independent of the API version and are now Apache 2.0 rather than MIT — worth a license review before upgrading. On the platform side, [Claude Models on Google Cloud](https://platform.claude.com/docs/en/release-notes/overview#claude-models-on-google-cloud-anthropic-managed-and-fast-mod) lets teams reach the Anthropic-operated API through a GCP project with IAM and Marketplace billing, with Fast mode for Opus 5.5; [organization web fetch controls](https://platform.claude.com/docs/en/release-notes/overview#organization-web-fetch-controls) add domain allow and block lists for admins; and [inference hooks graduated from beta](https://platform.claude.com/docs/en/release-notes/overview#user-profiles-organization-webhooks-and-inference-hooks-out-) alongside `/v1/user_profiles` and org webhooks. Governance news: Anthropic is [donating Agent Skills to the Agentic AI Foundation](https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation), now integrated by more than 70 agent platforms. And Claude Duet, a [proactive-agent app in research preview](https://support.claude.com/en/articles/12138966-release-notes#claude-duet-research-preview-for-pro-and-max-plans) for Pro and Max plans, is the week's most consumer-facing bet, padded out by [two](https://apps.apple.com/us/app/claude-by-anthropic/id6473753684?v=1.261023.20) [routine](https://apps.apple.com/us/app/claude-by-anthropic/id6473753684?v=1.261021.51) app maintenance releases.

### Workspace controls across ChatGPT plans

ChatGPT's business tiers converged on the same features: [locked chats on Business](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes#introducing-locked-chats) and [locked chats on Enterprise and Edu](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#introducing-locked-chats) hide conversations behind a PIN or device auth without breaking retention policies, Enterprise gained [Record for meetings and voice notes](https://help.openai.com/en/articles/11487532-chatgpt-record) and [shareable workspace profiles](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#share-profiles-within-your-enterprise-workspace), and Business admins got [member management in Admin Console](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes#manage-business-workspace-settings-in-admin-console). Enterprise analytics were [reorganized into Usage and Insights](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#workspace-analytics-updates-in-admin-console), with [a 7-day migration grace period](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#7-day-grace-period-for-personal-workspace-migration) for users leaving personal workspaces. Meanwhile [ChatGPT Ads expanded to 26 more countries](https://openai.com/index/chatgpt-ads-expands-26-new-countries).

## Releases covered

### Anthropic

- [Claude Code v2.1.309 fixes sandbox escapes and macOS startup-file leak](https://code.claude.com/docs/en/changelog#2-1-309)
- [Claude Code v2.1.305 fixes Vertex and Bedrock requests failing on tool search rejection](https://code.claude.com/docs/en/changelog#2-1-305)
- [Claude Code 2.1.304 fixes worktree removal losing commits and adds auto-mode setup](https://code.claude.com/docs/en/changelog#2-1-304)
- [Claude Code v2.1.307 adds /claude-api api-version and safety_event telemetry](https://code.claude.com/docs/en/changelog#2-1-307)
- [Claude Code v2.1.306 fixes cloud session login and spurious prompt-too-long errors](https://code.claude.com/docs/en/changelog#2-1-306)
- [Claude Code v2.1.310 fixes background cleanup deleting recent files](https://code.claude.com/docs/en/changelog#2-1-310)
- [Claude Code v2.1.308 bug fixes and reliability improvements](https://code.claude.com/docs/en/changelog#2-1-308)
- [Claude Platform ships API version 2026-10-21 with breaking SDK and CLI changes](https://platform.claude.com/docs/en/release-notes/overview#2026-10-21)
- [Claude Platform launches Anthropic-managed Claude Models on Google Cloud with Fast mode for Opus 5.5](https://platform.claude.com/docs/en/release-notes/overview#claude-models-on-google-cloud-anthropic-managed-and-fast-mod)
- [Claude Platform adds organization web fetch controls and domain restrictions](https://platform.claude.com/docs/en/release-notes/overview#organization-web-fetch-controls)
- [Claude Platform graduates Inference hooks from beta, launches user profiles and org webhooks](https://platform.claude.com/docs/en/release-notes/overview#user-profiles-organization-webhooks-and-inference-hooks-out-)
- [Anthropic donates Agent Skills open standard to Agentic AI Foundation](https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation)
- [Anthropic launches Claude Duet research preview for Pro and Max plans](https://support.claude.com/en/articles/12138966-release-notes#claude-duet-research-preview-for-pro-and-max-plans)
- [Claude by Anthropic 1.261023.20](https://apps.apple.com/us/app/claude-by-anthropic/id6473753684?v=1.261023.20)
- [Claude by Anthropic 1.261021.51](https://apps.apple.com/us/app/claude-by-anthropic/id6473753684?v=1.261021.51)

### OpenAI

- [OpenAI API adds 12-hour prompt caching option](https://developers.openai.com/api/docs/changelog#2026-10-20)
- [OpenAI API ships GPT-6.1 Sol GA and Decisions API GA with gpt-6-luna](https://developers.openai.com/api/docs/changelog#2026-10-22)
- [OpenAI launches Decisions API for fast, low-cost app decisions](https://openai.com/index/decisions-api)
- [OpenAI Go SDK v3.86.0 makes decision Instructions a typed union](https://github.com/openai/openai-go/releases/tag/v3.86.0)
- [OpenAI Node SDK v7.43.0 makes Decisions question instructions a string or parts array](https://github.com/openai/openai-node/releases/tag/v7.43.0)
- [OpenAI Python SDK v3.37.0 expands Decisions question and input types](https://github.com/openai/openai-python/releases/tag/v3.37.0)
- [OpenAI Node SDK v7.41.0 adds Vertex AI support for the Responses API](https://github.com/openai/openai-node/releases/tag/v7.41.0)
- [OpenAI Python SDK v3.36.0 adds Vertex AI support for Responses API](https://github.com/openai/openai-python/releases/tag/v3.36.0)
- [OpenAI Go SDK v3.85.0 adds Vertex AI support for the Responses API](https://github.com/openai/openai-go/releases/tag/v3.85.0)
- [OpenAI Go SDK v3/vertex v1.0.0 adds Vertex AI support for Responses API](https://github.com/openai/openai-go/releases/tag/v3/vertex/v1.0.0)
- [OpenAI Node SDK v7.40.0 adds voice consent management and cache-write usage counts](https://github.com/openai/openai-node/releases/tag/v7.40.0)
- [OpenAI Go SDK v3.84.1 fixes X.509 HTTP/2 transport race and WebSocket restore leak](https://github.com/openai/openai-go/releases/tag/v3.84.1)
- [OpenAI Go SDK v3.84.0 adds custom voice consent management](https://github.com/openai/openai-go/releases/tag/v3.84.0)
- [OpenAI Go SDK v3.83.0 adds raw WebSocket helpers and 30-minute cache-write usage counts](https://github.com/openai/openai-go/releases/tag/v3.83.0)
- [OpenAI Go SDK v3.85.1 respects per-attempt WebSocket auth timeout](https://github.com/openai/openai-go/releases/tag/v3.85.1)
- [ChatGPT Business adds locked chats hidden from history and search](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes#introducing-locked-chats)
- [ChatGPT Enterprise and Edu adds locked chats with PIN or device authentication](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#introducing-locked-chats)
- [ChatGPT Enterprise & Edu adds ChatGPT Record for meetings and voice notes](https://help.openai.com/en/articles/11487532-chatgpt-record)
- [ChatGPT Enterprise & Edu adds workspace profile sharing with owner controls](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#share-profiles-within-your-enterprise-workspace)
- [ChatGPT Business adds Admin Console workspace settings management](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes#manage-business-workspace-settings-in-admin-console)
- [ChatGPT Enterprise & Edu reorganizes workspace analytics in Admin Console](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#workspace-analytics-updates-in-admin-console)
- [ChatGPT Enterprise and Edu add 7-day grace period for personal workspace migration](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes#7-day-grace-period-for-personal-workspace-migration)
- [ChatGPT Ads expands to 26 new countries as 100,000 advertisers join](https://openai.com/index/chatgpt-ads-expands-26-new-countries)