ℹ️ Quick Answer: Nobody is winning the AI race outright in September 2026. OpenAI’s GPT-6 Astra tops the benchmarks and runs neck and neck with Claude on coding, while Anthropic is growing revenue fastest. Google’s Gemini and ChatGPT have the most users, and Meta’s new Muse is the hottest free app.
📋 WHAT’S INSIDE
- Who’s Winning the AI Race on Benchmarks?
- Which AI Is Best for Coding Right Now?
- Which AI Do the Most People Actually Use?
- Who’s Slipping?
- Frequently Asked Questions
Last updated September 27, 2026.
Three of the big labs shipped new flagship models inside the same week this month, and then Meta surprised everyone and took the number one spot on the App Store. So if you’re asking who’s winning the AI race right now, the honest answer depends on what you mean by winning.
I’m an engineer, and I pay for both sides of this fight. I’m moving off Claude’s biggest plan, Max 20x, down to Max 5x plus a ChatGPT Pro plan for Codex, which is OpenAI’s coding tool. ChatGPT and Claude are so close now that we’re the ones who benefit. There’s no reason to be locked into one brand when several of them do the same job.
Who’s Winning the AI Race on Benchmarks?
OpenAI, by a nose. On BenchLM’s September leaderboard, GPT-6 Astra scores 88.69, with Anthropic’s Claude Opus 5.5 right behind at about 87 and Claude Fable 5.1 at 83.
The timing was wild, because Anthropic released Fable 5.1 on September 1, Google put out Gemini 3.8 Flash the next day, and OpenAI answered with Astra on September 3. Then on September 22 OpenAI added GPT-6 Sol and Luna, and Anthropic launched Opus 5.5, which it says performs at Fable 5.1’s level on most work while costing 40 percent less to run than Opus 5. Gemini 3.8 Flash lands lower on that chart at about 73, and it’s built to be the cheap, fast option rather than the smartest one.
Which AI Is Best for Coding Right Now?
It’s a two-horse race between Anthropic and OpenAI. OpenAI calls Astra its “best model for software engineering to date,” and Anthropic says Opus 5.5 beat Fable 5.1 on agentic coding, meaning the AI works through a coding task on its own, in its own tests.
The money says Claude Code is doing fine. Anthropic’s annualized revenue passed $65 billion by late July according to Bloomberg and Axios, up from about $9 billion at the end of 2025, and Claude Code became the fastest-growing product in its history. The way I see it, for coding and for all-around usefulness it’s a close race between Claude Code and OpenAI’s Codex and ChatGPT, and those two are the leaders right now. I’m a big fan of the Sol and Astra models, and they’re a big part of why I think OpenAI and Anthropic are running neck and neck. I also still feel like ChatGPT and Codex give you more for what you pay, which is why I’m splitting my plans instead of picking one. If you want the details on what Opus 5.5 changes for paying users, I broke that down in this Opus 5.5 post.
Which AI Do the Most People Actually Use?
ChatGPT and Gemini, and it’s close. OpenAI’s last official count was 900 million weekly users in February, and Google says the Gemini app passed 1 billion monthly users in August. Weekly and monthly users aren’t the same measurement, so don’t read those as a head-to-head score.
The surprise is Meta. Its Muse app launched September 8 and hit number one on Apple’s free iPhone chart on September 18, Bloomberg reported, and downloads have since passed 3.4 million. It’s still tiny next to ChatGPT’s user base, but I’d call it a real contender. Muse is Meta, and Meta isn’t afraid to throw money at something new and push it. I’ve been seeing Muse ads everywhere I go, so if nothing else people are going to know about it, and marketing is usually the hardest part. When a company that big is willing to move fast to grab market share, or just create a whole new market, I take it seriously. For everyday use, I’d say ChatGPT is probably still leading for now, but Gemini is catching up for sure, because so many people already have a Gmail or Google account. Gemini can reach across your Google stuff natively, so it basically builds you a workspace that can see everything and recommend things. ChatGPT gets there more through third-party connectors. If you’re trying Muse, here’s how Muse stacks up against ChatGPT.
Who’s Slipping?
xAI is the one falling behind the pack. Grok 4.7 came out September 21 at the same price developers paid for Grok 4.6, and it improved on coding tests, but it scores 68 on the same BenchLM chart where Astra sits near 89. I haven’t played with Grok much, honestly, because I haven’t needed to.
Keep in mind that one leaderboard is one opinion, and the margins are so thin that Anthropic can lead today and a week from now it could be Grok or anyone else. I don’t think brand loyalty is really a thing here. People want the best models and the best output, and they don’t care much where they get it.
Frequently Asked Questions
Who is winning the AI race right now?
No single company. In September 2026 OpenAI’s GPT-6 Astra leads BenchLM’s overall leaderboard, Anthropic leads on revenue growth, Google’s Gemini app has over 1 billion monthly users, and Meta’s Muse is the top free app.
Which AI is best for coding in 2026?
It’s between OpenAI and Anthropic. OpenAI calls GPT-6 Astra its best software engineering model yet, and Anthropic says Claude Opus 5.5 beat Fable 5.1 on agentic coding.
Is Grok falling behind?
On benchmarks, yes. Grok 4.7 scores 68 on BenchLM’s September leaderboard, well below GPT-6 Astra and Claude Opus 5.5, though it costs developers less than the top models.
So it’s still very much an open race. I’ll update this in December with whoever’s on top by then.
Related reading: GPT-6 Sol and Luna explained | ChatGPT’s $500 plan | New to AI? Start here
WHO WROTE THIS
Moses Smith. I write Everyday AI for people who aren’t engineers. I go try the tools, then tell you honestly whether they were worth it. Sometimes the answer is no, and that’s kind of the point.
This blog is free and has no ads. If it saved you some time, you can buy me a coffee.
Leave a Reply