OpenAI Codex engineer Thibault Sottiaux posted a short update on X yesterday. Three bullet points. The last one said "Go do things." The message was clear: we're getting out of your way.
First: the 5-hour usage limit for Plus, Business, and Pro plans is temporarily gone. No end date announced. Second: GPT-5.6 Sol is getting efficiency improvements. Less usage consumed per task. Exact impact still being quantified. Third: active users hit 6 million. A usage reset is rolling out now.
The timing matters. Anthropic just delayed Fable 5's full rollout again. They extended the subscription access period for the seventh time. Users are frustrated. Some are canceling. OpenAI's response came within an hour. This is not a coincidence. This is a war. Not for benchmarks. For users.
Anthropic is rationing access because Fable 5 is compute-hungry and expensive to run. OpenAI is removing limits because it just made Sol cheaper to operate. The gap between these two strategies is now visible in real time.
OpenAI has been preparing for this moment. GPT-5.6 Sol already outperforms Fable 5 on the Artificial Analysis Coding Agent Index — 80 vs 77.2 — while using less than half the output tokens, taking half the time, and costing about one-third less. Sam Altman told CNBC that Sol is 54% more token-efficient on agentic coding tasks.
Now they're pushing that efficiency advantage even further. The exact numbers aren't public yet, but the direction is clear. When Anthropic has to tighten belts, OpenAI is loosening them. Users notice.
The 6 million user milestone tells the rest of the story. Axios reported in early June that Codex weekly active users had just crossed 5 million — a sixfold increase since the desktop app launched in February. A little over a month later, it's 6 million. 1 million net new users in about five weeks. The adoption curve is accelerating.
The 5-hour limit removal is a user acquisition play disguised as a customer service move. Remove friction when the competitor adds it. It's classic product strategy. And it works because the switching cost is low.
But here's the catch: "temporarily" is doing a lot of work. OpenAI could reinstate limits if Sol's efficiency gains don't keep pace with demand. The reset is a bridge, not a solution.
The efficiency improvements are the actual answer. Lower token consumption per task means the same subscription goes further. That's what keeps users on the platform after the reset buzz fades. The 54% efficiency gain Altman cited is the real moat.
OpenAI is betting on a simple equation: cheaper inference → more usage → more data → better models. Anthropic is betting on a different one: better reasoning → higher value → willing to pay more. Both could be right. But right now, OpenAI is winning the volume game.
If Anthropic keeps delaying Fable 5's full rollout, users will keep migrating. OpenAI's job is to make sure the migration path is as frictionless as possible. Removing the 5-hour limit is exactly that. It says: come over here. There's no line.
P.S. If you're a Codex user, you just got upgraded for free. In OpenAI terms, "temporarily removing the limit" means "we haven't figured out how to charge you for this yet." Enjoy it while it lasts.
