Skip to content
Daily Brief News Daily Brief

6 AI stories from June 14, 2026: OpenAI retires GPT-5.2, acquires Ona for Codex, Mistral eyes 3B euro raise, ChatGPT model picker simplified, Codex DevTools access, and Fable 5 still offline

· by Pondero Newsdesk · 6 stories

AI news daily brief: 2026-06-14

Six stories today: three OpenAI product moves (a silent GPT-5.2 retirement, a cloud-execution acquisition, and new Codex developer tooling), a major ChatGPT UX change, a Mistral fundraising report, and a day-two update on the Anthropic model suspension.

OpenAI silently retired all GPT-5.2 models on June 12 and migrated every user to GPT-5.5

As of June 12, 2026, GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro are gone from ChatGPT. Per OpenAI's release notes, existing conversations that used GPT-5.2 automatically continued on the corresponding GPT-5.5 model. The move followed OpenAI's 90-day successor policy, which had been triggered when GPT-5.3 Instant launched, but an August 10 deprecation date still listed in OpenAI's deprecations docs led many developers to believe they had weeks of runway remaining.

Canvas is also gone from GPT-5.5 Instant and GPT-5.5 Thinking; writing and coding functionality moved into inline writing blocks and code blocks. The retirement went live without a separate announcement post. Developers in the OpenAI community forum flagged the discrepancy between the model-picker disappearance and the docs page deadline. Full story: OpenAI GPT-5.2 retirement

OpenAI is acquiring Ona (Gitpod GmbH) to let Codex run long-horizon tasks unattended in the cloud

OpenAI announced on June 11 that it will acquire Ona, the German cloud-execution startup formerly known as Gitpod GmbH, to bring persistent cloud infrastructure into the Codex ecosystem. The deal has not yet closed; terms were not disclosed. Per OpenAI's announcement, the Ona team will join the Codex group after closing to build customer-controlled environments for agents that run "over hours or days, rather than minutes."

More than 5 million people use Codex each week, up 400% from earlier this year, per OpenAI. The acquisition addresses a structural gap: agentic coding tasks that span multiple sessions require persistent compute that outlasts a single browser tab. Bloomberg and CNBC both confirmed the deal. Full story: OpenAI acquires Ona for Codex

Mistral AI in talks to raise 3 billion euros at a 20 billion euro valuation, sources say

TechCrunch reported on June 12 that Mistral AI is in early discussions to raise approximately 3 billion euros at a valuation of roughly 20 billion euros, citing people familiar with the matter. Yahoo Finance confirmed the figures independently. Mistral's last disclosed round valued the Paris-based company at 11.7 billion euros in its September 2025 Series C, when ASML Holding took an 11% stake. The new round would represent a 71% valuation jump and bring total capital raised to roughly 7 billion euros.

Discussions are early and terms could change. Mistral has not confirmed the talks. Full story: Mistral 3B funding round

ChatGPT simplified its model picker: Instant, Medium, and High replace the confusing Thinking-tier labels

OpenAI updated the ChatGPT model picker on June 10 to use four simplified tiers: Instant (fast, light), Medium, High, and Extra High (Pro subscribers only). Thinking Light has been removed. Per OpenAI's release notes, GPT-5.5 models remain the underlying engine across all tiers.

An auto-switch option lets Instant promote itself to Medium for complex tasks; users can toggle it in General Settings. The new picker appears at the top of the conversation on iOS and Android, and in the message composer on web. The change collapses what had grown into an unwieldy list of named variants into a four-step effort dial. Full story: ChatGPT model picker simplified

OpenAI Codex gets Chrome DevTools access in Developer mode and referral-based rate-limit banking

OpenAI's June 11 Codex release added Developer mode for the Codex Browser on Chrome. When toggled in Codex app settings, the agent gets controlled Chrome DevTools Protocol access: JavaScript performance profiling, console output and network traffic inspection, DOM state review, applied styles, and live browser diagnostics. Per OpenAI's Codex changelog, the same release shipped referral invitations.

Eligible Plus and Pro users received one free rate-limit reset at launch. Both the sender and a new recipient get a banked reset when the recipient sends their first Codex message; banked resets are valid for 30 days. An /init command for generating an AGENTS.md scaffold also shipped in this release. Full story: Codex DevTools and referral banking

Fable 5 and Mythos 5 remain offline one day later: what enterprises should actually do when their AI model vanishes overnight

Claude Fable 5 and Mythos 5 remained suspended worldwide as of June 13, 2026, one day after the Commerce Department export directive forced Anthropic to pull both models globally. No emergency court stay has been filed, and Anthropic said only that it is "working to restore access as soon as possible," per the company's blog post.

VentureBeat published enterprise guidance advising teams that relied on Fable 5 to roll back to Opus 4.8 as a stop-gap and audit which internal tools assumed access to a specific Anthropic model tier. Fortune noted that because Anthropic cannot filter foreign nationals from US users in real time, a single government directive can force a full global blackout. The episode has enterprise risk managers asking whether AI vendor contracts should include "model continuity" SLAs.

Sources