releases.sh

BetaWeekly digests are a beta — we're trying something new. Feedback welcome.

Price cuts, Fast mode, and the end of GPT-5.4 in Codex

July 27 – August 2, 2026

OpenAI
OpenAI

OpenAI slashed GPT-5.6 Luna pricing by 80%, added Fast mode to the Sol API, and set an August 31 retirement date for GPT-5.4 models in Codex, while the SDKs gained new fast tier and content provenance support.

Dramatic pricing shifts and a faster Sol

The week's biggest story came from OpenAI's price cuts and Fast mode launch: GPT-5.6 Luna dropped 80% in price and GPT-5.6 Terra fell 20%, reflecting efficiency improvements the company previewed earlier. At the same time, the GPT-5.6 Sol API gained Fast mode, delivering up to 2.5× faster speeds at twice the price, replacing the Priority Processing system with backward compatibility. The SDKs followed quickly — Python SDK v2.51.0, Node SDK v7.2.0, and Go SDK v3.48.0 all added support for the fast tier.

Codex model retirements and the sign-in bet

Codex is retiring GPT-5.4 and GPT-5.4 mini for ChatGPT-signed-in users on August 31, with recommended replacements of GPT-5.6 Terra and GPT-5.6 Luna. The models remain available via the API and API-key-authenticated Codex sessions, but users need to update workspace defaults before the cutoff. Meanwhile, Sign in with ChatGPT launched in beta across select plugins and partner sites including Airtable, GitLab, HubSpot, Notion, Supabase, and Vercel — sharing only name, email, and profile picture. The Codex team rolled out the same feature for plugin connections, paired with a substantial Codex update adding multi-repository review, image editing, and browser history search. The Codex CLI 0.146.0 also shipped session naming, thread forking, and Agent Plugins marketplace support for Amazon Bedrock and Claude Code.

SDK maintenance: version bumps and provenance

The SDK teams pushed two major version floor updates. Python SDK v2.49.0 now requires Python 3.10 (up from 3.8), and Node SDK v7.0.0 requires Node.js 22 with a codified version support policy. Later in the week, Python SDK v2.52.0, Node SDK v7.3.0, and Go SDK v3.49.0 all added content provenance checks — a new API feature for verifying the authenticity and origin of AI-generated content. The Python update also improved Retry-After handling up to two minutes.

Enterprise controls and academic workspaces

ChatGPT Enterprise & Edu gained per-environment Work Local/Cloud controls, Fast Mode defaults with custom-role overrides, and a redesigned Roles and Permissions page. Separately, ChatGPT launched dedicated workspaces for academic researchers — faculty and postdoctoral researchers can apply for 12 months of free access for up to five verified team members, with business data protections and Pro-level usage limits.

AI-generated digests may contain mistakes.
Releases covered16