releases.shpreview

Home/Collections/Frontier AI Labs

Frontier AI Labs

Frontier AI research labs and model providers, side by side.

Show prereleases

.json·.md·.atom

MonJun 22, 20262 releases

Daily summary

Claude gains MCP CLI auth and bash auto-respond

Claude Code now lets users authenticate MCP servers directly from the command line and automatically responds to bash commands for a more conversational workflow.

Claude Code adds <code>claude mcp login</code> and <code>logout</code> for authenticating MCP servers from the CLI, with <code>--no-browser</code> support for SSH sessions
Bash <code>!</code> commands now trigger an automatic Claude response, with a setting to restore the previous context-only behavior
The desktop Claude app received a minor bug-fix release with no feature changes

Claude Code

MCP auth from CLI; bash commands auto-respond

v2.1.186

Added claude mcp login and logout for authenticating MCP servers from the CLI without the interactive menu, with --no-browser for SSH. Bash ! commands now trigger Claude to respond automatically; set "respondToBashCommands": false to restore the previous context-only behavior. Over 20 fixes including streaming requests failing after wake-from-sleep, subagent transcript scroll bleeding, and agent team effort-level inheritance.

Claude

Claude by Anthropic 1.260618.0

v1.260618.0

Squashed some bugs and improved the overall experience. Yours, Claude

Claude by Anthropic 1.260618.0

SunJun 21, 20261 release

Daily summary

ChatGPT auto-converts long pastes to attachments

ChatGPT now automatically converts pastes over 10,000 characters into attachments for Free and Go users, while raising the threshold to 10k for all paid plans.

ChatGPT now automatically turns long pastes into attachments across all plans, keeping the composer clean and preventing large text from consuming the context window.

ChatGPT

Large pastes become attachments on Free and Go plans

Long pastes over 10,000 characters are now automatically converted to attachments in ChatGPT for Free and Go users, keeping the composer cleaner and preventing large pastes from consuming the context window. The threshold was also raised to 10k characters for Plus, Pro, and Business users, up from the previous 5k limit.

SatJun 20, 20261 release

Daily summary

Claude Code adjusts stall messaging

Claude Code's stream-stall hint now appears less alarming and waits longer before showing, triggering after 20 seconds of silence instead of 10

Claude Code rewords its stall hint to "Waiting for API response · will retry in …" and doubles the quiet-period threshold before it appears

Claude Code

Stream-stall hint triggers after 20s with clearer wording

v2.1.185

The stream-stall hint now reads "Waiting for API response · will retry in …" instead of "No response from API · Retrying in …", and triggers after 20 seconds of silence instead of 10.

ThuJun 18, 20266 releases

Daily summary

Frontier labs tighten enterprise controls and deliver health intelligence

Anthropic shipped Claude Code artifact sharing and enterprise-managed MCP auth while OpenAI launched GPT-5.5 Instant with improved health responses and ChatGPT Enterprise credit analytics.

Claude Code blocks destructive git and terraform destroy commands unless explicitly requested, and warns when a model has been auto-upgraded or deprecated
Claude Code Artifacts beta turns raw session output into shareable visual pages for Team and Enterprise users
Anthropic lets admins centrally authorize MCP connectors through an identity provider, removing per-user setup
GPT-5.5 Instant now matches frontier Thinking models on health evaluations and reduced factuality errors by 71%, available to all free ChatGPT users
ChatGPT Enterprise admins get credit usage analytics across users and models, plus default and per-user spend limits

Claude Code

Destructive git commands blocked; fullscreen TUI corruption fixed

v2.1.183

Destructive git commands (git reset --hard, git checkout --., git clean -fd, git stash drop) are now blocked when you didn't ask to discard local work. Fixed fullscreen TUI corruption in Windows Terminal under heavy nested-subagent load, blank turns from thinking-only model responses, and MCP servers exposing auth-stub tools in headless mode.

Product Launches

2 updates

Artifacts beta: raw output becomes shareable visual pages

Artifacts in Claude Code convert raw output into visual pages that can be shared with team members. Available in beta on Team and Enterprise plans.

Artifacts in Claude Code: share your work as it happens

MCP connectors authorized through identity providers

Admins can now authorize and authenticate MCP connectors across their organization through an identity provider, removing individual setup.

Enterprise-managed auth for MCP connectors

Product Releases

2 updates

Credit usage analytics and granular spend controls for Enterprise

ChatGPT Enterprise admins can now view credit usage trends across users, products, and models in the Global Admin Console, and set default, group-specific, or individual spend limits. Users can see their credit usage and request limit increases with context.

Billing tab and overall plan view in the global admin console

GPT-5.5 Instant health performance matches frontier models

GPT-5.5 Instant now performs at a level comparable to frontier Thinking models on health evaluations, available to all free ChatGPT users. Production factuality issues flagged in health responses have fallen by 71% in the last two months, and physician-rated responses now surpass both older models and human-written responses across accuracy, communication, and completeness criteria.

Go SDKv3.41.0

WedJun 17, 20267 releases

Daily summary

Claude Code adds prompt-level config, OpenAI ships spend alerts

Claude Code gained a /config syntax for setting any option from the prompt and fixed prompt caching on custom base URLs, while OpenAI's Go SDK exposed admin spend alerts via API.

Claude Code now accepts /config key=value from the prompt to set any setting, and prompt caching works correctly on custom ANTHROPIC_BASE_URL and Foundry endpoints
OpenAI Go SDK added the admin organization spend alerts API

ChatGPT Business

Record & Replay turns workflows into reusable skills

Record & Replay is a new Codex app feature for macOS that lets eligible Business users demonstrate a workflow once and turn it into a reusable skill. Initial availability excludes the European Union, UK and Switzerland.

Codex

2 updates

Record & Replay turns workflows into reusable skills

Record & Replay lets macOS users turn a demonstrated workflow into a reusable skill, initially available outside the EEA, UK, and Switzerland with Computer Use enabled. Bulk actions for automation run history, deep links for SSH connections, and improved Browser Use session persistence are also included.

Remote executor channels now e2e encrypted; WAL-reset corruption fixedv0.141.0

Remote executors now use authenticated, end-to-end encrypted Noise relay channels, and bundled SQLite is pinned to a version containing the WAL-reset corruption fix. Plus cross-platform remote execution, per-thread executor plugin MCP servers, and auto-resolving TUI prompts.

Python SDKv2.43.0

Node SDKv6.44.0

Go SDKv3.40.0

Claude Code

Prompt cache now works on custom URLs; Bun 1.4 bundled

v2.1.181

Prompt caching now reads correctly on custom ANTHROPIC_BASE_URL and Foundry settings, among dozens of other fixes. A new /config key=value syntax lets you set any setting from the prompt, and the bundled Bun runtime is upgraded to 1.4. Plus fixes for network drive writes, macOS TUI freezes, and startup blocking.

TueJun 16, 20263 releases

Daily summary

Claude Code keeps partial streams; OpenAI adds spend alerts

Claude Code gained resilience against dropped connections by preserving partial responses, while OpenAI shipped API-level admin spend alerts in its Python SDK.

Claude Code now retains partial responses if a connection drops mid-stream, preventing lost work and fixing a stuck spinner, WSL2 scrolling, and Bash-tool sandbox issues.
OpenAI Python SDK added an API endpoint for admin spend alerts, enabling programmatic budget monitoring.

Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem

Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem.

ChatGPT Enterprise & Edu

Edu workspace members can export data when enabled by admin

Members of ChatGPT Edu workspaces without data residency configuration can now export workspace data when an admin enables the Data export setting. Exports are requested from Settings > Data controls and delivered via email as conversation JSON files.

Python SDKv2.42.0