Did OpenAI quietly cut GPT-5.6 Sol reasoning? FixlationAI said yes. Tibo Sottiaux said no — and shipped ~10% more Sol usage from inference savings the same morning.
On July 13, 2026, @thsottiaux (OpenAI Codex) posted a five-thread update wave for ChatGPT Work and Codex users: banked resets scaling to 7M active users, a Sol subscription commitment, and a direct rebuttal to thinking-budget nerf rumors. Same weekend Anthropic extended Fable 5 through July 19 — two labs racing on quota psychology, not just benchmark points.
TL;DR — what people are asking
Question
Answer (July 13)
Did Sol get nerfed?
Tibo: no · FixlationAI: yes — reasoning down one tier
@FixlationAI posted that OpenAI reduced GPT-5.6 Sol thinking budgets to make the model more efficient:
"They essentially bumped everyone's reasoning down by 1... so if you were running Sol Extra High, you now have to set it to Max to get the same effort."
That matches developer anxiety after Tibo's July 12 efficiency rollout — Sol tasks were already burning less weekly quota, which felt like less reasoning even when framed as optimization.
Tibo counter (~10 minutes before this post)
@thsottiaux replied in a thread titled "Updates for Codex and ChatGPT Work users. No nerfing, only good stuff!":
Claim
Tibo's response
Reasoning cut
"No nerfing, only good stuff"
Efficiency changes
Inference optimizations landed — savings passed to subscriptions
Net quota impact
~10% more GPT-5.6 Sol usage on its own
Effort tiers
Tweet continued (truncated in live thread) — framed as passing savings down, not lowering ceilings
explainx.ai read: When a lab ships efficiency patches during a limit crisis week, users interpret token savings as quality cuts. Tibo's 10% usage bump is the explicit counter-narrative — more runway, not less brain.
If you run xhigh or Extra High on long Codex loops (Matt Shumer's week-long Manhattan build), verify your effort picker after each Tibo deploy. Community reports on effort drift are anecdotal until OpenAI publishes tier token budgets.
For Claude-side effort semantics, see effort parameter guide — Sol's tier names differ but the quota-per-reasoning-level tradeoff is the same product problem.
Banked reset rollout — 500k today, 7M tomorrow
July 13 — infrastructure + bug fix
Tibo's 5-hour-ago thread on banked resets:
Item
Detail
Scope
500k ChatGPT Work + Codex users
New surface
Web + mobile banked reset (desktop-only before)
Bug
<10% of users who triggered reset in a 2-hour window did not actually refill
Remediation
Banked reset granted to everyone who pressed reset in that window
Purpose
Validate infra before fleet-wide grant
July 14 — milestone grant
Tibo previewed tomorrow:
7M active users celebration
First banked reset across all ChatGPT Work and Codex users
Two hours before the nerf thread, Tibo posted a subscription commitment after launch-week churn fear:
"Rest assured that GPT 5.6 Sol will stay in the ChatGPT subscription you pay for. Including Go, Plus and Pro subscriptions. At least until we ship an even better model."
Follow-up: Team, Edu, and any other paid subscription — same rule.
That matters because:
July 12 temporarily removed the 5-hour gate — users worried the promo was a bait-and-switch before repricing
Same morning, @claudeai (9h earlier): "We're extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code's weekly rate limits 50% higher, through July 19."
Effort-tier and banked-reset details reflect Tibo's July 13 posts as of publication. Verify quota and effort settings in your account before production sprints — launch-week policy changes without always publishing rollback dates.
