Codex cuts Fast mode multiplier to 2x
OpenAI updated Codex pricing documentation to lower the Fast mode quota multiplier for GPT-6.2 Astra and Sol from 2.5x to 2x. Ultrafast mode usage also dropped from 8x to 6x, significantly reducing quota burn for low-latency coding workflows.
Lowering Fast mode multipliers signals that OpenAI's serving costs for the GPT-6.2 architecture are dropping, making high-speed coding sessions far more economical for developers.
- –Included usage for Fast mode on GPT-6.2 Astra and Sol now consumes subscription quota at 2x instead of the 2.5x penalty still applied to GPT-6 and GPT-6.1.
- –Ultrafast mode sees a larger drop from 8x to 6x for included allowance, with pay-as-you-go overages temporarily discounted to 3x through February 2027.
- –High multipliers previously pushed developers to reserve Fast mode for urgent debugging; narrowing the gap encourages keeping accelerated generation on by default.
- –By tying the cheaper speed multipliers exclusively to GPT-6.2, OpenAI creates a direct incentive to migrate agentic workflows off legacy models.
- –With rivals like Cursor and Claude Code aggressively optimizing turn latency, reducing inference friction in Codex helps OpenAI defend its developer mindshare.
DISCOVERED
9h ago
2026-11-22
PUBLISHED
9h ago
2026-11-22
RELEVANCE
AUTHOR
AdamHoltererer