Wednesday, September 30, 2026

TLDR

OpenAI DevDay delivered: GPT-6.1 Sol at $2/$10 per M tokensA small unit of data, such as part of a word, that AI models read and generate. Many AI services price by tokens in and out.NVIDIA ↗ (~1/5 the cost of GPT-6 Astra), always-on "Dots" agents, Agents API with computer use, and an "Ultrafast" speed tier (Sep 29) Pricing Agents
Trump and AI leaders signed a voluntary, "morally binding" safety accord after the White House lunch — no enforcement mechanism (Sep 29) Policy
OpenAI is reportedly seeking $30B at a ~$1.4T valuation; IPO pushed to 2027 (Sep 29) Trending
Gemini gemini-omni-flash-preview shuts down today (Sep 30) Pricing

Things your engineering team should look at or do this week.

Benchmark gpt-6.1-sol against Claude Sonnet 5.5 on your own coding/agent tasks — both list at $2/$10 per M tokens; OpenAI also lists $0.10 cached input. Compare cost per completed task, not list price (vendor benchmark claims only).
Confirm nothing in your stack still calls gemini-omni-flash-preview (shuts down today) and review the Oct 1–2 deprecations listed below.
Evaluate the new Agents API / Codex Security Cloud in a sandbox before any production use; apply the egress-allow-listing guidance from yesterday's sandbox-escape findings to anything with computer-use access.

GPT-6.1 Sol (Sep 29) Pricing — OpenAI says it roughly matches GPT-6 Astra on its coding benchmark (DeepSWE v1.1) at about one-fifth the cost, and beats Claude Opus 5.5 on AutomationBench at ~1/3 the cost (OpenAI's own numbers). Available as gpt-6.1-sol; $2 input / $0.10 cached / $10 output per M tokens. (OpenAI)
Dots: always-on agents (Sep 29) Agents Customer AI — Persistent agents that work across ChatGPT, text, email and Slack with cloud computer and browser access; available on Pro and Business Premium, beta for Enterprise/Edu/Healthcare. (OpenAI)
Codex updates: Codex in the Cloud, refreshed CLI, automatic GitHub/GitLab code review, Codex Security Cloud (Sep 29) — Remote dev environments, a CLI with voice steering and /agents view, and repo scanning with threat investigation. Plan availability varies. (OpenAI)
Ultrafast tier (Sep 29) Pricing — Up to ~8x faster generation in Codex and ~6x in the API on Pro 500/Enterprise; GPT-6.1 Sol Ultrafast is "coming soon." (OpenAI)

OpenAI Agents API, Decisions API, Bedrock Managed Agents (Sep 29) Agents — Agents API adds computer use, multi-agent orchestration and context compaction; Decisions API is in limited preview; OpenAI agents will also run AWS-native via Bedrock. (OpenAI)
ChatGPT Pro 500 ($500/month) and "Sign in with ChatGPT" (Sep 29) Pricing — 25x Plus allowance plus Ultrafast access; Devin, Notion, Vercel and others will accept ChatGPT sign-in and plan usage. (OpenAI)

OpenAI reportedly seeking $30B+ at ~$1.4T valuation (Sep 29) Trending — Follows a $122B round at $852B in March; IPO now expected in 2027, with Altman ruling out a 2026 listing. Reported, not closed. (TechCrunch)

White House AI accord: voluntary "statement of principles" (Sep 29) Trending — Trump, Amodei, Zuckerberg, Huang, Musk and others signed a joint commitment to robust internal controls on "super intelligence"; Trump called it "morally binding" and "almost like a constitution," but it is voluntary with no stated enforcement. Limited detail available at publication. (CBS News)
OpenAI calls for mandatory, capability-based federal AI safety rules (Sep 9, context for the accord) — Its policy post urges Congress to require testing, cyber protections and incident reporting rather than rely on voluntary commitments. (OpenAI)

Gemini gemini-omni-flash-preview shutdown — today, Sep 30.
OpenAI gpt-5.4-cyber removed — Oct 1; GitHub Copilot billing changes Oct 1, model retirements Oct 2.
NYC Council AI hearing — Oct 5; UK CMA choice-screen consultation closes — Oct 9.
GPT-5.5 retired from ChatGPT — Oct 14; OpenAI o1/o3-mini/o4-mini API shutdown — Oct 23.
Haiku 5.5 — "coming weeks"; Cursor loses OpenAI model access — Nov 12; Gemini Gems → Skills — Nov 17.