DevDay 2026 Recap
OpenAI's DevDay 2026 introduced over 20 announcements, including a new Decisions API built on Luna, Codex updates, and controversial subscription tier changes.
summary is enough The official post lacks technical depth, but the comment thread provides sharp insights on the new Decisions API and pricing backlash.
summary
OpenAI's DevDay 2026 featured over 20 updates across ChatGPT, Codex, and agent tools targeting 1.2 billion weekly users. The most discussed technical addition among developers is the Decisions API running on the Luna model, designed for high-volume classification and business automation. Meanwhile, users heavily criticized abrupt subscription adjustments that restructured top-tier usage limits and pricing.
key points
- OpenAI introduced a Decisions API powered by the Luna model to handle high-volume classification and business automation tasks.
- Commenters strongly criticized subscription changes that effectively halved usage limits for existing high-tier plans while pushing previous limits into a new $500 tier.
- Additional releases included Codex security scans and sub-account sign-in functionality.
who should read it
Developers and technical leaders tracking OpenAI's product roadmap, pricing shifts, and API offerings.
discussion skeptical
consensus
Commenters agree that the Decisions API is a major focus for business automation, but express widespread frustration over subscription price hikes and skepticism toward novelty features.
pushback
Some commenters noted that the rapid rollout of the Decisions API effectively undercuts independent competitors in the classification space.
No one mentioning Decisions API?
Feels huge. Sounds like Luna on their Ultrafast infra.
I did some internal benchmarking and found that GPT 6 Luna in batch mode at 20 records per batch was about 1.6x faster at 1.2x the cost of Jev with no tradeoff in accuracy. Tuning it up/down would make it faster per-record while also reducing the cost (assuming accuracy holds). Though this was only useful for offline processing.
It shares early internal benchmark data comparing Luna's batch mode throughput and cost against alternative models.
Generated by an LLM from the article and thread. Check the sources before relying on it.