Onstage at TechCrunch Disrupt 2026, Replit CEO Amjad Masad said the company was no longer working with Anthropic, then clarified on X on Oct 17 that Replit remains model-agnostic: users can still pick Anthropic and other models, but OpenAI is now the preferred partner powering Replit Agent by default, citing performance, cost and partnership. Since most users keep the default, the change is likely to shift a large share of Replit spend between the two labs.
Key Takeaways
- ✓Masad clarified on Oct 17: OpenAI is the preferred partner powering Replit Agent; Claude and other models remain selectable
- ✓Masad says Sol and Astra score better on Replit's internal and public benchmarks for building apps from scratch; no scores disclosed
- ✓Per TechCrunch, Replit was until recently a top-five Anthropic customer, and a late-September blog still listed Claude Fable 5/5.1 in Agent production
- ✓Masad says enterprise contracts are ~15% of revenue this year (vs ~2% last year) and B2B is over 50%
- ✓A person familiar with the Anthropic side says the partnership is still in place; Anthropic has not commented officially
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Core Background & Industry Pain Points
Replit is a browser-based AI coding platform popular with founders and non-specialists; Replit Agent turns a prompt into a deployed app. Per TechCrunch, Replit long relied on Anthropic's Claude and was until recently a top-five Anthropic customer, before announcing an OpenAI partnership in August. For such platforms the default model drives both user experience and margins, because most users never change it.
Architecture Highlights & Internals
What changed is the default routing, not model availability. Masad stated on X that Replit stays model-agnostic and users can still choose Anthropic and other supported models; OpenAI is now the preferred vendor behind Replit Agent's default experience. Replit's docs describe a model selector plus Free, Auto, Power and Max modes, with Auto picking a model per paid task.
Authoritative Benchmarks & Measured Scores
Masad said Replit's internal benchmark and its public benchmark consistently show Sol and Astra performing better at building vibe-coded apps from scratch, but he gave no scores and we found none; treat this as a vendor claim with no published numbers. A person familiar with the Anthropic side says the partnership is still in place; Anthropic has not commented officially.
Developer Hands-on Guide
Existing Replit users may see new tasks default to OpenAI models, with changes in output style, speed and billing. Claude fans can switch in Agent's model selector and compare results on the same prompt; workspace admins should review model controls. For other AI coding tools, it shows defaults increasingly follow cost per task.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.