Feature Request: Dedicated Premium Requests for GPT-6 Astra with Auto-Fallback to GPT-5.6 Sol on 0%

ChatGPT Feature requests
, ,
1

Summary

Currently, running GPT-6 Astra (especially with Fast Mode or Ultra reasoning) rapidly consumes the rolling 5-hour usage allowance in ChatGPT Work and Codex—frequently dropping remaining quota from ~96% to ~6% in a single execution. When Astra hits 0%, active development hits a hard stop.

Proposed Architecture

  1. Dedicated Premium Request Pools: Replace the rolling 5-hour window for heavy reasoning with a transparent pool of Premium Requests / Weekly Allowance reserved exclusively for GPT-6 Astra.
  2. Automatic Fallback to GPT-5.6 Sol: When the Astra/Premium allowance reaches 0%, the active session or task should automatically downshift to GPT-5.6 Sol (or standard execution) rather than locking the user out of Codex and ChatGPT Work completely.
  3. Graceful Subagent Scaling: Under the GPT-5.6 Sol fallback state, multi-turn tasks and agent runs should continue executing without hard-stopping background work.

Why This Improves Developer Experience

  • No Sudden Interruptions: Developers aren’t left stranded midway through a refactor or code review when Astra finishes its allocation.
  • Predictable Budgeting: Users can consume Astra Premium requests deliberately for high-judgment tasks, knowing their baseline coding velocity remains intact on GPT-5.6 Sol.
1 Like