ARTICLE AD BOX
OpenAI just dropped the five-hour usage cap on its top-tier GPT-5.6 Sol model, and if you’re using ChatGPT Plus, Pro, or Business, you can enjoy that right now. The change comes after a rush of users piled on after July 9, when Sol rolled out to a wider audience.
People jumped on the chance to try the new model—developers, businesses, AI hobbyists, pretty much everyone who follows this stuff. That demand pushed OpenAI to loosen the restrictions, at least for now, as it works on beefing up its servers to keep up.
Sol’s had a strong start. It shot to the top of the Design Arena leaderboard with an Elo score of 1,353, which says a lot about how people rate its reasoning and coding chops.
OpenAI gave a handful of partners early access on June 26, then opened the doors to the public a couple weeks later. CEO Sam Altman said they talked things out with the U.S. government ahead of launch—Altman called it a “collaborative back and forth” with the Trump administration.
So, what’s in the GPT-5.6 family? There are three flavors:
Sol – The heavy hitter. It’s made for complex code, tough research, and anything enterprise-level.
Terra – The middle ground. It handles everyday pro tasks without breaking a sweat.
Luna – The fast, affordable one. Perfect for when you need a lot of responses at a lower cost.
OpenAI says Sol is a beast with tokens—54% more efficient on tricky agent-based coding work. That means you get more done with fewer tokens and save money if you’re running big jobs.
On benchmarks, Sol is turning heads. It scored 80 on the Coding Agent Index. For context, that’s nearly three points higher than Anthropic’s Fable 5, and Sol spits out results in half the time, at about a third of the cost, with way fewer tokens. On “Agents’ Last Exam,” which covers 55 job categories, Sol scored 53.6—more than 13 points above Claude Fable 5.
Microsoft’s all in on this, too. The company announced that GPT-5.6 is taking over as the main AI brain for Microsoft 365 Copilot—so it’ll power Word, Excel, PowerPoint, Chat, and Cowork. Microsoft and OpenAI go way back, and this move doubles down, even with rumors swirling that Microsoft’s building some of its own AI models to cut costs.
Let’s talk money. Here’s the new pay-as-you-go setup for the GPT-5.6 models:
Sol: $5 for a million input tokens, $30 for a million output tokens
Terra: $2.50 input, $15 output
Luna: $1 input, $6 output
All three have a knowledge cutoff in February 2026, support a one-million-token context window, and can spit out up to 128,000 tokens per request.
You can try these models through ChatGPT, the OpenAI API, Codex, and GitHub Copilot, giving everyone—including enterprise developers—a bunch of ways to get access.
The temporary lift on Sol’s usage limits shows OpenAI wants people to have a smooth ride as it rolls out one of its biggest models yet. But look, Sol’s premium price tag probably means most users stick with Terra or Luna for everyday stuff, saving Sol for when things get serious—advanced coding, automation, or big enterprise workflows.
The pressure’s on OpenAI to juggle hardware, prices, and safety as demand picks up. For Microsoft, putting GPT-5.6 at the heart of its productivity tools might push even more businesses to jump on board, as long as they can keep a handle on costs, oversight, and how they use AI responsibly.
English (US) ·