Replit CEO Amjad Masad said during an onstage interview at TechCrunch Disrupt 2026 that the company is no longer working with Anthropic, framing the decision around cost, partnership, and performance. The session recording was published by TechCrunch on October 16, 2026.
Onstage Confirmation and Rationale
The exchange began when the interviewer asked whether the OpenAI partnership Replit announced in August 2026 meant the company was no longer working with Anthropic. “Yeah. I mean, yeah,” Masad answered.
Masad said OpenAI’s models now perform better for Replit’s workloads. “I think GPT is better now,” he said, citing an internal Replit benchmark called VBench and a public benchmark the company has published. He said OpenAI’s Sol and Astra models consistently performed better at building vibe-coded applications from scratch, and he described the switch as a decision based on cost, partnership, and performance, with OpenAI the better partner for now.
When the interviewer asked whether losing Replit had contributed to what he described as Anthropic’s reported revenue slowdown, Masad said Replit was also one of Anthropic’s top five customers.
Earlier in the session, asked whether Replit is vulnerable to the same disruption it creates as OpenAI and Anthropic move beyond code generation into building, deploying, and hosting apps, Masad said those offerings remain surface level next to Replit’s full-stack, integrated environment, noting that OpenAI relies on an external service, Cloudflare, for hosting. He said Replit’s agents can spin up databases and key-value stores inside a production environment and that the company can deploy its entire stack into government and enterprise environments as an on-premises solution.
Claude Models, Benchmarks, and the OpenAI Partnership
Replit’s current model-selector documentation lists Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5, Claude Opus 5, Claude Opus 5 Fast, Claude Opus 4.8, and Claude Sonnet 4.6 as Max Mode options, alongside GPT-6 Astra, GPT-6 Luna Fast, and Kimi K3, with GPT-6 Sol offered in Power Mode. Free Mode always uses intelligent model routing with no manual selection, and the Fast variants require a Pro or Enterprise plan.
A September 29, 2026 engineering post on Replit’s blog documented Claude Fable 5 running in Replit Agent production in August 2026 and Fable 5.1 alongside GPT-6 Astra in September 2026. The post’s production data shows Fable 5 dispatching a subagent on 32% of turns, Fable 5.1 on 21%, and Astra on 36%, each at medium reasoning effort.
The post reports that Replit Agent, evaluated in Max mode with Astra as its core loop, scored 72% at $2.11 per task on the DeepSWE v1.1 benchmark and 49% at $2.53 per task on Terminal-Bench 4.0, beating a sidekick architecture, the same setup with one long-lived worker, by 11 and 16 points. Replit’s engineers wrote that the code quality of Astra and Fable 5.1 let the company use its code-review subagent less with no drop in evaluation scores, describing it as a level of engineering quality they had not seen from any model before.
The OpenAI partnership itself dates to Replit’s August 18, 2026 announcement of Free Mode, powered by OpenAI’s GPT-5.6 Luna. Replit said Core subscribers could create up to 30 times more than before for their $20 monthly subscription, including up to 30 hours per month of chat, with everyday tasks in Free Mode no longer consuming credits. The announcement describes Power Mode, formerly Economy Mode, as using cost-optimized models and Max Mode as using higher-performance models, with usage limits resetting every five hours.
Enterprise Mix and Financial Posture
In the same session, Masad said enterprise contracts account for about 15% of Replit’s revenue in 2026, versus about 2% in 2025, and that more than 50% of revenue is now B2B. He declined to share ARR figures.
Masad said companies that adopt Replit spend 400% more within 12 months, a net revenue retention rate he put above 400%, and said Replit has been gross-margin positive with margins growing alongside enterprise adoption.
Masad also described what he called an agent habitat of orchestration, error recovery, sandboxing, networking, and identity and authorization as the prerequisite for deploying AI safely inside organizations. He outlined Replit’s pre-deployment security scans, post-deployment penetration-test scans, supply-chain monitoring, and a one-click kill switch that lets enterprise customers take down any app or all apps. He referenced Replit’s September 2024 launch of Agent 1 and a 2025 incident in which an agent deleted a database, saying stored snapshots make restoration a one-click action.
Masad said there are no immediate plans to raise new funding, that an IPO is in the company’s future, and that Replit has turned down acquisition approaches because it is focused on building a long-term company.