ChatGPT $100-Tier User: GPT-6 Astra Burns 3x Usage Budget Yet Codes Worse Than GPT-5.6 Sol
F0xy1337 · reddit · 2026-09-16
A Reddit user who upgraded to ChatGPT's $100 plan for GPT-6 Astra reports that for programming it consumes usage far faster while producing less practical progress than GPT-5.6 Sol High — citing overengineering and long unproductive discussions. After burning roughly three separate 5× usage budgets experimenting with skills, MCP project context, and different workflows, he concludes Sol delivers more useful output per unit of usage, and questions the price-to-value ratio for software development.
More from coding & agent
- OpenAI launches Data agent in ChatGPT Work to query company data and build dashboards — gdb · 2026-09-16
- Muse agent renews passport and books flights end-to-end in viral demo — armand_ruiz · 2026-09-16
- Bitsec's multi-model agent stack found 160+ exploits, beating a single 'superhuman' model — markjeffrey · 2026-09-16
- Dev slams agent sandbox pricing as 20x+ the cost of a $6/month always-on VPS — Aryvyo · 2026-09-16
- Abacus.AI says Smaug Flash fixes open-source models' tool-call hangs in production — bindureddy · 2026-09-16
- Dev lets AI agents reverse engineer Zapier and build his own automation app — NathanWilbanks_ · 2026-09-16