Replit CEO claims that the company no longer cooperates with Anthropic
TechCrunch
4h ago
Ai Focus
Replit CEO, co-founder with Amjad Masad, stated on TechCrunch Disrupt that the company has ceased cooperation with Anthropic due to reasons including cost, partnership, and performance. He mentioned that Replit still allows users to choose the Anthropic model, but the ChatGPT of OpenAI is now the default option; at the same time, the proportion of the company's corporate business is on the rise.
Helpful
No.Help

The CEO and co-founder of this programming platform, which is highly popular among startup founders and developers, stated on Thursday at the event that the company will no longer cooperate with [partner name].

Masad indicates that this decision was based on various reasons, including cost. Replit announced in August that it had reached a partnership with OpenAI, but at that time, there were no signs that it would no longer cooperate with Anthropic.

“I think ChatGPT is better now,” said Masad, adding that the company uses internal benchmarks to evaluate performance and other indicators. “We have also released public benchmarks, and throughout the process, Sol and Astra have performed better in building vibe-coded applications from scratch. This decision is based on cost, partnership, and performance. For all these reasons, we simply feel that they are the better partners at this time.”

However, Anthropic does not seem to be completely excluded by Replit. Replit sells a cloud-based software development platform, where users can directly develop and deploy applications in their browsers.

After the Disrupt event concluded, Masad further clarified that customers can still access the models on the platform, but ChatGPT of OpenAI is now the default option. This change may significantly reduce Replit's expenses on Anthropic, as most customers typically stick to the default settings rather than manually selecting models themselves.

Before reaching an agreement with OpenAI, Replit relied on Anthropic's Claude model to operate its platform. It is said that this collaboration helped the company experience significant revenue growth. Masad stated on stage that until recently, Replit was still one of Anthropic's top five customers.

“Replit is still model-independent,” Masad said to TechCrunch on the weekend. “Our users can choose from Anthropic’s Claude or other supported models… Based on performance, economy, and the strength of our partnership, OpenAI is now our preferred partner for providing support for Replit Agent.”

However, insiders familiar with the cooperation relationship between Anthropic and Replit raised objections to the claims made by Masad on stage. A person with knowledge of the matter, who requested to remain anonymous, stated that this cooperation relationship still exists.

Company websites, blog articles, and social media posts also show that Replit was still promoting the products of Anthropic until just a few weeks ago. For example, in a blog post at the end of September, Replit mentioned that Anthropic's Claude Fable 5 and Claude Fable 5.1 were models running in the Replit Agent production environment. Around the same time, a product employee from Replit posted on X stating that they were using Anthropic's flagship model Opus 5.5 and promoting its capabilities.

Masad also shared information about the company's growth at the Disrupt event, stating that the business has begun to account for a larger share of revenue.

"We are not yet releasing the data for ARR, but in terms of growth, I would say that last year corporate business accounted for about 2% of our revenue," said Masad. "This year, I would estimate that around 15% of our revenue comes from corporate contracts, and more than 50% of our revenue comes from B2B."

Masad told TechCrunch in private that by 2027, more than half of Replit's revenue is expected to come from corporate and government clients.

Masad also discussed the necessity of good AI security measures.

“Just as we humans need a certain environment to work – you need a good office, perhaps also good food, air conditioning, a computer – intelligent agents also need a place to reside in order to exist and perform their tasks well,” he said. “Therefore, this environment requires proper organization, robust error recovery, strong security and protection measures, as well as sandbox mechanisms, a reliable network, and proper identity and authorization systems. Once you have all these in place, you can securely deploy AI.”

This article was initially published on October 16th at 2:22 PM Pacific Time, and later updated to provide more background information about the collaboration between Replit, Anthropic, and OpenAI.

Tip
$0
Like
0
Save
0
Views 47
HQYC reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
Trump claims he "can do the impossible," and during his campaign in Kentucky, he stated that if Republicans win, the Department of Defense will revitalize a local pharmaceutical factory
On Saturday, while campaigning for Republican candidates in Kentucky, Trump boasted about his power in the White House, stating that if the Republicans won, the Department of Defense would invest in reviving a local pharmaceutical factory; he also delivered a series of speeches on tariffs, the Iran war, oil prices, and congressional control.
Fortune
·2026-10-18 05:47:14
29
Meta Removes Clauses That Once Caused Concerns About Web Content Caching; AI Addresses Training Issues
Meta recently updated its terms of commercial use, stating that the company may use enterprise account domain names, web pages, or any content from URL to support, develop, improve, research, train, and maintain the integrity of Meta products and services. Legal experts interviewed by Nieman Lab believe that this wording may grant Meta the authority to scrape any publisher's website that has an Meta enterprise account for use in training AI models. However, a spokesperson for Meta stated that the updated terms do not imply any change in the way Meta scrapes online content. Subsequently, The Verge reported that this term has now been deleted.
TechCrunch
·2026-10-18 05:39:50
19
Pump.fun Launches Perpetual Contract Service, Initiating Cryptocurrency Leverage Trading
Pump.fun launches perpetual contract services, bringing Bitcoin and Ethereum leveraged trading to the platform through Hyperliquid, with maximum leverage of up to 40 times for Bitcoin and 25 times for Ethereum respectively. Hyperliquid is responsible for execution, margin requirements, funding rates, clearing, and settlement, while Pump.fun only provides the trading interface; this service is not available to users in the United States, Ontario, and other restricted regions.
The Cryptonomist
·2026-10-18 05:17:02
24
Ledger Hardware wallet tampering raises security concerns for 2026
The CEO of Mt Gox stated that reports of tampered hardware wallets can be traced back to at least December 2025, and questioned the company for not issuing more proactive warnings to customers regarding known risks. Security researcher Joe Grand also mentioned that he reported an issue with a Nano X device purchased through a Thai Lazada to Ledger in March 2026. Ledger indicated that an internal investigation has been launched regarding these findings, and in another incident involving CryptoBilis, it was confirmed that some devices contained unauthorized embedded components.
The Cryptonomist
·2026-10-18 05:16:59
23
Anthropic claims its employee structure is less "junior-oriented," and meeting times are also increasing
Anthropic released a report on Friday stating that the company's employee structure has become less "junior-oriented" over the past year, with the proportion of technical teams declining, while meeting times have increased. The company also mentioned that the adoption of AI has improved productivity, but "human attention and judgment" are becoming bottlenecks.
TechCrunch
·2026-10-18 04:37:14
30
View More