Anthropic’s Claude Code is holding its lead in the developer tools race, even as OpenAI’s Codex and a wave of open-source coding models pull more attention. That’s the read from The Information, which reports that despite rising interest in rival options, Claude Code remains the tool developers reach for first. What stands out here isn’t that Anthropic is winning. It’s that it’s winning the one AI category where switching costs are supposed to be low and competition is fierce.
Coding is the beachhead. It’s the clearest example of AI agents doing real, paid work today, and every major lab knows it. That’s why this matters now.
What’s actually changing
The competitive field has gotten crowded fast:
- OpenAI’s Codex has been rebuilt as an agentic coding tool and is drawing serious developer interest, according to The Information.
- Open-source models aimed at code keep closing the quality gap, and they come with a pitch enterprises love: run it yourself, keep your data, skip the per-seat bill.
- Anthropic’s Claude Code sits on top anyway, driven by model quality and a workflow developers trust for large, messy codebases.
So you’d expect the leader to be sliding. Instead, developers are experimenting with alternatives while still doing their real work in Claude Code. Curiosity and commitment aren’t the same thing.
Why the lead is holding
A few dynamics explain it. Model quality still decides who wins on the hardest tasks, and Anthropic’s models have earned a reputation for handling long context and complex refactors without falling apart. Habit matters too. Once a coding agent is wired into someone’s daily flow, the bar to rip it out is high. “Good enough” from a competitor rarely clears it.
There’s also a revenue story. Coding tools are one of the few AI products generating real, recurring income right now, which is exactly why OpenAI is pushing Codex so hard and why open-source backers keep shipping. Anthropic reportedly draws a meaningful chunk of its revenue from this category. Defending it isn’t optional for them.
The Future Cast
Here’s where this goes over the next 12 to 24 months. Expect the gap on raw capability to keep narrowing. Open-source code models will get good enough for a large share of everyday tasks, and the fight shifts from “whose model is smartest” to “whose tool fits my workflow, my security rules, and my budget.”
That means Anthropic’s moat becomes less about the model alone and more about integration, reliability, and trust. The risk for any leader in a fast-moving category is complacency. Interest in Codex and open-source isn’t noise. It’s the early signal of buyers keeping their options open.
What to do about it
If you build software or run a team that does, a few practical moves:
- Don’t marry one tool. Keep Claude Code if it’s your workhorse, but pilot Codex and an open-source model on a real project. You want to know the gap firsthand, not from a benchmark chart.
- Weigh data control, not just quality. For sensitive codebases, a self-hosted open-source model may beat a smarter cloud tool on the factors your security team actually cares about.
- Track cost per developer. As open-source options mature, the price of proprietary agents becomes a live negotiation, not a fixed line item.
- Watch the switching cost you’re building. The deeper a tool sits in your workflow, the harder it is to leave. That’s fine if it’s the best tool. It’s a trap if you stopped checking.
Anthropic is winning the coding agent race for now, and the reasons are real. But “for now” is doing a lot of work in that sentence. The interesting question isn’t whether Claude Code leads today. It’s whether Anthropic can turn a model-quality lead into a durable one before the rest of the field catches up. More details are in the original report from The Information.