Eight million active users across Codex and ChatGPT Work — in roughly five months from ~1 million in February.
Update — July 29, 2026: Another Work/Codex reset — Sol efficiency ~18%, five-hour limit returns Jul 30. Full update.
Update — July 23, 2026: Tibo is now teasing a Codex reveal for July 24. See what is confirmed, what developers are guessing, and what to verify tomorrow.
On July 14, 2026, Tibo Sottiaux (@thsottiaux, OpenAI Codex) posted the milestone, reset usage limits again, and confirmed the 5-hour cap stays off so weekly quota governs Plus, Pro, and Business agent sessions. Sam Altman (@sama) called GPT-5.6 Sol growth "insane", cited a 2.5× spike in agentic product usage over the prior week, and warned inference scaling hiccups may follow.
The story is not just a user number — it is OpenAI buying retention with quota gifts while GPT-5.6 Sol (July 9 GA) meets a merged desktop harness (Chat / Work / Codex) that confused part of the installed base. explainx.ai maps growth math, limit policy, and what subscribers should expect next.
TL;DR
| Metric | Detail |
|---|---|
| Active users (Tibo, Jul 14) | 8M combined Codex + ChatGPT Work |
| Prior milestone | 6M on July 12 |
| Feb 2026 baseline | ~1M (community / prior Tibo posts) |
| Recent jump | ~6M → 8M in ~3 days (X commentary) |
| Usage spike | 2.5× agentic usage week-over-week (Altman) |
| Limit policy | Another full reset · no 5h window (still temporary) |
| Model | GPT-5.6 Sol in Go, Plus, Pro subscriptions |
| Risk | Altman: scaling hiccups possible despite inference heroics |
| Grok headline | 7M weekly — may differ from Tibo's active count |
Growth timeline — February to 8M
| Date | Milestone | Source |
|---|---|---|
| Feb 2026 | ~1M active users (Codex + Work trajectory) | Community recap / launch baseline |
| Jul 9 | GPT-5.6 Sol / Terra / Luna GA | OpenAI launch |
| Jul 10–11 | Multiple quota resets · traffic "unmatched" | Tibo launch-week thread |
| Jul 12 | 6M active · 5h cap removed · weekly 100% reset | Tibo |
| Jul 13 | Banked resets rollout · 7M preview · Sol stays in paid subs | Tibo + thinking-budget post |
| Jul 14 | 8M active · another reset · no 5h limit continues | Tibo + Altman |
explainx.ai read: The steepest segment is post-Sol GA — not gradual organic growth. 6M → 8M in ~72 hours (per @haider1 and launch-week X) implies promo + product merge + limit gifts converting lurkers into agentic sessions, not only new installs.
What Tibo announced at 8M
"Hello. We have reached 8M active users across Codex and ChatGPT Work. We are once again resetting the usage limits for all. And we continue to not have the 5h rate limit as well, allowing everyone to explore the boundaries of GPT-5.6 Sol and discover how ambitious you can be." — Tibo Sottiaux, @thsottiaux, July 14, 2026
| Action | Meaning for subscribers |
|---|---|
| Full usage reset | Weekly (and related) meters topped up again — manual gift, not plan tier change |
| No 5-hour limit | Long Sol Ultra / multi-step Work runs governed by weekly cap only — still temporary |
| Sol exploration | OpenAI wants ambitious agent loops while infra catches up |
Tibo teased "8M celebration day" the evening before — the post delivered the number plus more quota.
Staged rollouts: Some free accounts and non-Pro tiers reported missing resets in Tibo's replies earlier in the week — same pattern as banked resets hitting 500k users before all users on July 14.
Altman on inference — heroics and hiccups
| Post | Quote / signal |
|---|---|
| ~Jul 14 | "5.6 sol growth is insane. the inference team has done heroic work to be able to support demand." |
| Same thread | "we are going to move mountains to continue to scale, but it is possible there are some hiccups soon." |
| Earlier week | "2.5x increase in usage of our agentic products (codex and chatgpt work) in the last week!" |
Translation for builders: Expect latency spikes, queued tasks, or surprise limit tightening if 2.5× load outruns GPU provisioned capacity — even with Sol efficiency patches Tibo shipped July 12.
Parallel week: Anthropic extended Fable promos and free Claude for Teachers — different retention levers, same frontier demand war.
Why usage exploded — product + policy stack
1. GPT-5.6 Sol as default agent brain
Sol GA July 9 put a frontier coding model inside the same subscription many users already paid for. Tibo committed Sol stays in Go, Plus, and Pro until something better ships.
2. ChatGPT desktop merge (Work + Codex)
The three-mode app expanded Codex beyond developers into deliverable Work — documents, slides, research — while sharing one agentic quota pool. Power users run terminals in Work and repos in Codex in the same week.
3. Limit policy as growth lever
Launch week interventions (timeline):
- Jul 10–11: Repeated 100% resets
- Jul 12: Remove 5h gate
- Jul 13: Banked resets + ~10% more usage from inference savings (Tibo)
- Jul 14: Another reset at 8M
Pattern: OpenAI trades margin for MAU and session depth during Sol trial — similar to Tibo's May/June 2026 emergency top-ups, but at 8× scale.
4. Automation stories on X
Developer posts cite end-to-end agent wins — e.g. wiring RevenueCat without manual SDK spelunking — that were impractical pre-Sol or pre-merge. Those anecdotes drive 2.5× usage more than homepage hero copy.
Friction still on the ground
| Complaint | Status |
|---|---|
| App merge confusing | Work vs Codex vs Chat switcher · shared history until separated |
| Cloud vs desktop sync | Work threads on web/mobile still don't sync to desktop at launch |
| Free tier resets | Lagging paid rollouts |
| "Resubscribe bait" skepticism | Users ask when 5h limit returns — Tibo never promised permanence |
| Capacity | Altman's hiccups warning |
Not the story: Unlimited frontier access. Weekly caps still bind — OpenAI is widening the pipe and refilling meters, not removing economics.
8M vs 7M — which number is right?
| Source | Figure | Metric |
|---|---|---|
| Tibo (Jul 14) | 8M | Active users · Codex + ChatGPT Work combined |
| Grok news summary | 7M | Weekly users (aggregated X narrative) |
| Tibo (Jul 12) | 6M | Active users |
explainx.ai guidance: Cite 8M active from Tibo for product milestones. Weekly active (WAU) can lag or differ from point-in-time active definitions OpenAI does not document publicly. Without an SEC filing or blog chart, treat all figures as directional, not audited DAU.
What developers should do this week
| If you… | Then… |
|---|---|
| Hit limits before | Check if weekly meter reset landed · confirm 5h row gone in UI |
| Run long Codex jobs | Batch work before potential hiccups · export artifacts locally |
| Pay for Pro/Plus | Sol is included — optimize prompts/harness, not model access |
| Need clarity on modes | Read Work vs Codex guide |
| Compare frontier labs | GPT-5.6 vs Fable · limit wars through Jul 19 |
Summary
Codex and ChatGPT Work hit 8 million active users on July 14, 2026 — up from ~1 million in February and 6 million two days earlier — powered by GPT-5.6 Sol, a merged agentic desktop app, and repeat quota resets with the 5-hour cap still removed. Sam Altman reported 2.5× agentic usage growth in a week and warned inference scaling hiccups despite heroic capacity work.
For subscribers, the actionable read is dual: use the promo window for ambitious Sol agent loops while resets hold, and don't treat 8M hype as permanent unlimited access — weekly governance and possible outages remain. For the market, 8M combined agentic MAU in five months is OpenAI's clearest signal yet that coding agents + Work deliverables are becoming a mainstream paid surface, not a developer preview.
Related Reading
- GPT-5.6 Sol usage ~18% longer + 5-hour returns (Jul 29)
- ChatGPT Work thread orchestration — mobile July 26
- Codex Micro — first OpenAI branded hardware (Jul 15, 2026)
- ChatGPT Drops 5-Hour Limit — Tibo Resets (Jul 12)
- ChatGPT Work vs Codex — Complete Guide
- GPT-5.6 Sol vs Claude Fable 5
- GPT-5.6 Sol Thinking Budget & Banked Resets
- Codex vs Claude Code Comparison
- Fable 5 Extended Through July 19
- OpenAI Codex Plugin for Claude Code
- macOS blocks Codex as malware — XProtect false positive fix (Jul 2026)
Tibo and Altman posts are accurate as of July 14–15, 2026 per public X threads. User counts are company-reported active users, not independently audited. Limit policies are temporary unless OpenAI states otherwise on openai.com or help.openai.com.