Update: DevDay is tomorrow — leaked plans put free API in Codex
OpenAI's Sept 29 event is hours away; three plans — Free, Prototype, Accelerate — leaked from its code. Plus frontier tiers from Z.ai and Qwen, callable now.
Copy markdown
The leaked tier trio: Free, Prototype, Accelerate
OpenAI's codebase reportedly holds three new plans: a Free tier with API access through Codex plus limited Playgrounds, a Prototype tier with higher limits across all models, and an Accelerate production tier starting at $50. Nothing is official until the keynote, but free Codex API access — if it ships — lets a solo builder prototype against OpenAI models without a card on file.
DevDay 2026 is locked for Sept 29
The event is confirmed for Tuesday, Sept 29, built around API and developer-tool sessions, with a global multi-city tour to follow. Model and pricing news tends to drop live — pin your current model IDs and note your rate limits today so you can diff exactly what changes the moment it lands.
GLM 5.3 Prime: same weights, faster tier
Z.ai's GLM 5.3 Prime is live at $2.80/$8.80 per million tokens with a 1M-token context — but it runs the same weights as GLM 5.3 on a lower-latency tier, not a smarter model. Reach for it when speed matters; don't pay the premium expecting better answers.
Qwen3.8 Max Prime lands at $4/$12
Alibaba's Qwen3.8 Max Prime ships a 1M-token context at $4/$12 per million tokens, tuned for agentic and coding work. It's the frontier-tier counterweight to GLM 5.3 Prime if you want a second closed model to A/B against inside your harness.