Replit Picks OpenAI Over Anthropic as Default [2026]

Replit spent a single week teaching the AI coding world a lesson about the gap between a stage soundbite and a company’s actual contracts. On October 16, 2026, Replit CEO Amjad Masad told a packed room at TechCrunch Disrupt that his company was “no longer working with Anthropic.” By October 17, he was publicly walking the line back, telling reporters that Replit “remains model agnostic” and that users could still choose Anthropic’s Claude inside the platform. What survived the clarification was simpler and arguably more interesting: OpenAI is now Replit Agent’s preferred, default AI partner.

The switch lands in the middle of the most competitive stretch the AI coding agent market has seen, with Claude Code and Cursor racing to add tools and GitHub Copilot and Cursor trading blows over which platform developers should build inside. Replit’s move does not end its relationship with Anthropic, but it does reset who gets top billing on one of the most visible AI coding products on the market, and it says a lot about where the economics of the agent business are heading in late 2026.

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

What Amjad Masad Actually Said at TechCrunch Disrupt

Masad’s original comment, delivered onstage at TechCrunch Disrupt on October 16, 2026, was blunt: Replit was “no longer working with Anthropic.” For a platform that had spent much of 2025 and early 2026 showcasing Claude-powered coding demos, that was a striking thing to say out loud, on the record, in front of an industry audience. It read, at first, like a full breakup between Replit and one of the two AI labs that had powered its flagship product.

Masad later told Digital Today that “I think ChatGPT is better now,” a comment that framed the switch as a straightforward quality call rather than a business dispute. That line, short as it is, is the closest thing to a plain-English reason Replit has given for why OpenAI now sits ahead of Anthropic inside Replit Agent.

The Clarification: “Replit Remains Model Agnostic”

The walkback came fast. Within a day, Masad posted a fuller explanation that softened the stage comment considerably. According to Digg, Masad wrote: “To clarify: Replit remains model agnostic. Our users can choose Anthropic models as well as many other supported models. What I meant at TC Disrupt is that OpenAI is now our preferred partner powering Agent based on performance, cost, and strength of our partnership.”

That is a meaningfully different claim than “no longer working with Anthropic.” It describes a default setting, not a severed relationship. IBTimes Singapore captured a near-identical version of the same clarification, quoting Masad saying “Our users can choose Anthropic’s Claude or other supported models” while OpenAI becomes “our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.”

Put the two statements side by side and the actual news becomes clear. Replit changed its default model recommendation. It did not remove Anthropic’s models from the product, and it did not announce the end of a contract.

Why Replit Says It Picked OpenAI

Masad’s clarification repeats three words every time he explains the decision: performance, economics, and partnership. None of those come with a published number. Replit has not released a benchmark score showing OpenAI’s models outperforming Claude on coding tasks, nor has it published a cost comparison showing how much cheaper the new default is to run. What it has done is tie the decision to its broader relationship with OpenAI, which deepened in August 2026 when the two companies struck a closer partnership agreement, according to reporting cited by The Beat.

That absence of hard numbers matters for how seriously outsiders should take the “performance” claim. Coding agent platforms live and die on model routing decisions, and a vague “it’s better now” from a CEO is not the same as a disclosed benchmark. Readers should treat the performance claim as Replit’s stated opinion rather than an independently verified result until the company, or a third party, publishes actual figures.

Is Claude Still Available on Replit? Yes, But It Is No Longer the Default

Multiple outlets confirm the same detail: Claude has not been pulled from Replit. Users can still select Anthropic’s models from the platform’s model picker. What changed is the out-of-the-box experience. New users and anyone who does not actively switch providers will now run Replit Agent on OpenAI’s models by default, rather than Anthropic’s.

Default settings matter more than they sound like they should. Most users of any software product never touch the settings menu. A default model change effectively redirects the bulk of day-to-day usage, and therefore the bulk of inference revenue, from one AI lab to another, even while leaving the technical option to switch fully intact. That is why Masad’s stage comment and his written clarification can both be technically accurate while landing very differently with an audience.

The August 2026 OpenAI Partnership Behind the Switch

The timing only makes sense once the August 2026 OpenAI-Replit partnership enters the picture. The Beat’s reporting ties the current default switch directly back to that earlier deal, describing it as the foundation for OpenAI’s new status as Replit’s “preferred partner.” Precise terms of that August agreement have not been made public, but the sequence suggests Replit and OpenAI had already been building toward a tighter integration months before Masad’s TechCrunch Disrupt comments forced the relationship into public view ahead of schedule.

That sequencing also explains why Replit’s position looked confused for about 24 hours. A partnership that had been negotiated quietly over months got compressed into a single, blunt soundbite onstage, and then had to be unpacked in public once Anthropic’s side of the relationship became a story in its own right.

Timeline: How the Story Unfolded in Four Days

Date	Event	Key Detail
August 2026	Replit and OpenAI deepen their partnership	Terms not disclosed; later cited as the basis for the default switch
Oct. 16, 2026	Masad speaks at TechCrunch Disrupt	Says Replit is “no longer working with Anthropic”
Oct. 17, 2026	Masad issues a public clarification	Says Replit is “model agnostic” and OpenAI is the “preferred partner”
Oct. 18 to 19, 2026	Coverage spreads across tech and business outlets	Anthropic’s own public-benefit fund news surfaces the same week

Why a “Preferred Partner” Switch Is Bigger Than It Sounds

It would be easy to read this as a minor settings change dressed up as news. The reason it is not minor is distribution. Replit Agent is one of the more visible consumer-facing entry points into AI-assisted coding, sitting alongside local agent tools like OpenCode and the big incumbent platforms. Every AI lab wants to be the default model behind the tools that millions of developers, students, and hobbyists open every day, because defaults compound. A developer who starts a project with one model’s quirks, context window limits, and code style tends to stay with it.

For OpenAI, becoming Replit’s preferred partner is a distribution win that does not require winning every individual developer’s trust one at a time. For Anthropic, losing default status on a platform it had been closely associated with is a visible setback, even if the underlying technical relationship has not actually ended.

The AI Coding Agent Market Has Never Been This Crowded

Replit’s reshuffle is landing in a market that is already saturated with competing claims about which model and which agent framework deserves developer loyalty. Anthropic has been pushing hard on its own tooling, recently expanding Claude Code’s plugin ecosystem past 300 tools while Cursor answers with its own environment features. Microsoft’s GitHub Copilot is simultaneously trimming its model lineup, a reminder that even the largest platforms are actively pruning which models they support rather than offering every option forever.

Pricing is part of that same fight. Anthropic has been aggressively cutting prices on smaller models to stay competitive on cost, a dynamic visible in Claude Haiku’s recent 90 percent price cut, while cross-lab pricing comparisons like the one between GPT-6.1 Sol, Sonnet 5.5, and Gemini 3.8 Flash show just how close the per-token economics of the major labs have gotten. Replit citing “economics” as one of its three stated reasons for the OpenAI switch fits a market where every serious AI lab is competing as hard on price as on raw capability.

Replit Agent Versus the Rest of the Field

Platform	Primary Model Relationship as of Oct. 19, 2026	Recent Headline
Replit Agent	OpenAI is the new default; Anthropic’s Claude remains selectable	Default switch confirmed Oct. 17, 2026
Claude Code	Anthropic’s own models, expanding third-party plugin support	Plugin ecosystem passed 341 tools
GitHub Copilot	Multi-model, but trimming its roster	Set to drop three AI models by Nov. 16, 2026
Cursor	Multi-model, competing on agent environments and tooling	Expanding environments API in direct response to Claude Code

Anthropic’s Side of the Story Surfaced the Same Week

Oddly, Anthropic’s own news cycle collided with Replit’s in the same stretch of days. The AI Insider reported on Anthropic detailing a shift in how it talks about AI’s effect on its own workforce, alongside the company’s separate commitment to steer a share of its equity into a public benefit fund, a move this site covered in detail earlier this week. None of that is directly caused by Replit’s default switch, but the overlap put Anthropic’s broader public posture, on both AI labor effects and its own corporate structure, in front of the same readers who were parsing Masad’s comments about preferring OpenAI.

It also matters that Anthropic has not issued a public rebuttal to Masad’s comments. Reporting from IBTimes Singapore notes that a source familiar with the Replit-Anthropic relationship disputed the idea that the partnership had ended at all, which lines up with Claude’s continued availability inside Replit. Anthropic choosing not to escalate the dispute publicly suggests the company sees this as a distribution setback worth absorbing quietly rather than a contract fight worth litigating in the press.

What Developers Are Actually Reacting To

The most consistent reaction documented across coverage is confusion rather than outrage. Developers who build on Replit were not angry that OpenAI became the default, according to The Beat’s account of the fallout. They were confused by the gap between Masad’s onstage statement and his written clarification a day later, which is a communications problem as much as a product one. A CEO saying a company is “no longer working with” a major AI lab, then clarifying that the lab’s models are still fully available, creates exactly the kind of whiplash that spreads across developer forums faster than the underlying facts can catch up.

That dynamic is familiar to anyone who has watched the AI coding tools market absorb a record quarter of venture funding while still struggling to communicate basic product changes clearly. Fast-moving platforms built on fast-moving model partnerships are going to keep producing these kinds of public corrections, because the underlying business relationships change faster than most companies’ communications processes can keep up with.

Historical Context: From Web IDE to AI Agent Platform

Replit built its early reputation as a browser-based coding environment aimed at students and hobbyists long before AI agents became the center of its product. Its pivot toward Replit Agent, an AI system that can plan, write, and run code with minimal human intervention, turned the company into one of the more visible consumer brands in the AI coding space, competing for attention against developer-first tools like Cursor and enterprise-first tools like GitHub Copilot.

That consumer visibility is exactly why Masad’s TechCrunch Disrupt comment traveled so fast. Replit is not a backend infrastructure company whose model choices stay invisible to end users. It is a platform where the model behind the agent is part of the product experience itself, which means every model decision Replit makes becomes a story in a way that a quieter infrastructure switch at a less visible company would not.

Market Impact: What This Means for OpenAI, Anthropic, and Replit

For OpenAI, the Replit default is a visible proof point in its broader push to be the backend of choice for agentic coding products, a push that overlaps with its other developer tooling investments. For Anthropic, the episode is a reminder that even close platform partnerships can get relabeled as “preferred” status shifts elsewhere, regardless of whether the underlying technical relationship survives intact. For Replit, the lasting effect may be reputational rather than technical: a company that markets itself as neutral and model agnostic just demonstrated, in real time, how quickly that neutrality narrative can get tangled by a single unscripted comment onstage.

Investors and enterprise buyers evaluating AI coding platforms will likely read this episode as a data point on vendor lock-in risk. If a platform’s default model can flip within a single news cycle based on a CEO’s offhand comment, buyers building critical workflows on any single model relationship have one more reason to insist on multi-model flexibility rather than betting on any one lab’s continued favor.

Expert Reactions and Industry Statements

Amjad Masad, Replit’s co-founder and CEO, has been the only named voice directly addressing the switch. Asked about his original TechCrunch Disrupt remark, he told Digital Today simply: “I think ChatGPT is better now.”

In his written clarification, reported by IBTimes Singapore, Masad laid out the fuller picture: “Our users can choose Anthropic’s Claude or other supported models…OpenAI is now our preferred partner for powering Replit Agent, based on performance, economics, and the strength of our partnership.”

He went further in a post captured by Digg, writing: “To clarify: Replit remains model agnostic. Our users can choose Anthropic models as well as many other supported models. What I meant at TC Disrupt is that OpenAI is now our preferred partner powering Agent based on performance, cost, and strength of our partnership.”

Taken together, the three statements tell a consistent story once the initial soundbite is set aside: a default change, not a breakup, driven by a mix of model quality judgment, cost considerations, and an existing partnership that predates the public comments by at least two months.

Predictions: Where This Story Goes From Here

Expect Replit to eventually publish, or leak, some version of a benchmark or cost comparison to back up the “performance” and “economics” language, since vague justifications tend to invite pressure for receipts once a switch becomes this visible.
Anthropic is unlikely to respond publicly in the short term, continuing the pattern of letting the dispute die down rather than amplifying it, especially while its own public benefit fund commitments dominate its press cycle.
Other multi-model coding platforms will face more pointed questions from users about which lab is their true default, now that the gap between “model agnostic” marketing and actual default settings has been exposed so clearly.
OpenAI will likely cite the Replit relationship in future developer-facing marketing as evidence of its growing footprint in agentic coding, alongside its other coding tool partnerships.
More CEOs at AI-native platforms will choose their words about model partnerships more carefully on conference stages, given how fast Masad’s comment traveled before his clarification could catch up.

What This Episode Says About the State of AI Coding Tools

Step back from the specific quotes and the Replit story is really a snapshot of where the AI coding agent market sits in October 2026. Every major platform, from Replit Agent to Claude Code to GitHub Copilot to Cursor, is negotiating the same tension between offering developers choice and quietly steering them toward whichever model relationship makes the most business sense that quarter. “Model agnostic” has become a popular phrase for platforms to use in their marketing, but this week showed how thin that label can be once a CEO says something plainly different in an unscripted moment.

None of that makes Replit’s technology worse, and it does not mean Claude’s coding performance has actually declined. It means the business layer sitting on top of these models, the partnerships, the default settings, and the public messaging around them, is now moving fast enough to generate its own news cycle independent of any actual model release.

Frequently Asked Questions

Did Replit actually stop working with Anthropic?

No. CEO Amjad Masad’s original comment at TechCrunch Disrupt said Replit was “no longer working with Anthropic,” but he clarified a day later that Claude remains fully available to users. OpenAI simply became the platform’s new default, preferred provider.

Why did Replit switch its default AI model to OpenAI?

Masad cited performance, cost, and the strength of Replit’s partnership with OpenAI. He has not published a benchmark or cost figure to back up those claims.

Can Replit users still choose Claude?

Yes. Claude and other supported models remain selectable inside Replit Agent’s model picker. The change only affects what new or default sessions use automatically.

When did Replit and OpenAI deepen their partnership?

Reporting ties the closer Replit-OpenAI relationship back to August 2026, months before Masad’s October comments made the arrangement public knowledge.

Has Anthropic responded to Replit’s comments?

Anthropic has not issued a public statement directly addressing the switch. A source familiar with the relationship reportedly disputed that the partnership had ended, which is consistent with Claude’s continued availability on the platform.

What OpenAI model does Replit Agent use by default now?

Available reporting confirms OpenAI as the new default provider but does not name a specific model version. Replit has not published that detail publicly as of October 19, 2026.

Does this affect other AI coding platforms like Cursor or GitHub Copilot?

Not directly. Replit’s default change is specific to its own platform. It does add pressure on competing platforms to clarify their own model relationships, given how much attention the Replit story drew.

Is this a sign that Claude’s coding performance is falling behind?

No independent benchmark has been published to support that conclusion. Masad’s comment that “ChatGPT is better now” is a stated opinion from one customer, not a verified, third-party performance result.

Replit’s Oct. 16 Anthropic Flip: OpenAI Now Default [2026]