+2 others
Anthropic's ChatGPT
Jan-Erik Asplund
TL;DR: OpenAI and Cursor have begun ganging up to challenge Anthropic's coding supremacy, but the launch of Claude Code has given Anthropic its own ChatGPT. Sacra estimates Claude Code hit $400M in annualized revenue in July 2025, up 23x over the past three months alone, generating 8% of Anthropic’s total $5B in annual revenue. For more, check out our full report and dataset.
Here’s our Anthropic update with key points via Sacra AI:
- With GPT-5, OpenAI is attacking Anthropic's supremacy in AI coding head-on with a model trained specifically for coding that 1) runs at 40-45% the average cost of Claude Sonnet 4 and 15% the cost of Opus 4.1 and 2) edges both on the SWE-bench coding benchmark. GPT-5 achieved a 74.9% on SWE-bench Verified, a benchmark that measures models' ability to solve real GitHub issues, versus Opus 4.1’s 74.5% and Sonnet 4’s 72.7%—on pricing, GPT-5 ($10 per 1M output tokens) is even cheaper than GPT-4o (launched May 2024), while Anthropic has kept the pricing on its Opus models at $75 per 1M output tokens since March 2024.
- At OpenAI’s Red Wedding aka the GPT-5 launch event, Anthropic partners Cursor ($500M ARR as of May 2025) and Jetbrains ($593M revenue in 2024) flipped default models from Sonnet 4 to GPT-5, while Vercel & Lovable have declared GPT-5 the new SOTA in coding. OpenRouter data shows Claude Sonnet 4 still dominates programming among end-users in bring-your-own-key AI IDEs like Cline and Amazon’s Kilo on token usage (361B tokens this week vs. 39.4B for GPT-5), demonstrating that Claude’s coating moat persists in tools where developers actively choose & pay for their models vs. using the default models in flat-rate products like Cursor.
- Anthropic's answer to this cutthroat defection lies in Claude Code, its ChatGPT: an owned, high engagement destination client vertically integrated with its models that grew a Sacra-estimated 23x in the last 3 months alone, from ~$17.5M in April 2025 to $400M in July. While OpenAI-aligned Cursor & Github fight over the IDE, Claude Code owns the terminal—a strategic bet on 1) the command line interface (CLI) as the most powerful interface for AI coding agents, and 2) the value of direct access to 100B+ daily code-edit tokens for RLHF.
For more, check out this other research from our platform:
- Anthropic (dataset)
- OpenAI (dataset)
- Vibe coding index
- Cursor (dataset)
- Cursor at $100M ARR
- Lovable vs Bolt.new vs Cursor
- Lovable (dataset)
- Bolt.new (dataset)
- OpenAI (dataset)
- Scale (dataset)
- Edo Liberty, founder and CEO of Pinecone, on the companies indexed on OpenAI
- Will Bryk, CEO of Exa, on building search for AI agents
- AI talking heads growing 1024%
- Kyle Corbitt, CEO of OpenPipe, on the future of fine-tuning LLMs
- How AI is transforming productivity apps
- AI and the future of video
Read more from
+1 others
Anthropic vs. OpenAI
Free Report
Continue Reading
OpenAI governance by force
Free Report
Continue Reading
OpenAI revenue, growth, and valuation
Unlocked Report
Continue Reading
Read more from
Cursor at $200M ARR
Free Report
Continue Reading
Cursor at $100M ARR
Free Report
Continue Reading
+2 others
Lovable vs Bolt.new vs Cursor
Free Report
Continue Reading
Read more from
Claude Code vs. Cursor
Free Report
Continue Reading
Vibe coding index
Free Report
Continue Reading
Anthropic at $316M ARR
Free Report
Continue Reading
Read more from
+1 others
Why OpenAI wants Windsurf
Free Report
Continue Reading
+2 others
OpenAI vs. Anthropic vs. Cohere
Free Report
Continue Reading
Anthropic revenue, growth, and valuation
Unlocked Report
Continue Reading
Read more from
Replit at $106M ARR
Free Report
Continue Reading
