IMPLICATOR
ChatGPT retook first in the Implicator LLM Meter on Aug. 17, ending Claude's one-week lead. ChatGPT scored 88 while Claude fell to 84, tied to an enterprise revenue crossover at OpenAI and Anthropic's no-opt-out watermark. Claude led for one edition, against at least two for every previous new leader.
The weekly editorial scorecard rates 10 models from 0 to 100 for enterprise buyers on compliance and security, professional-task quality, reliability and service commitments, ecosystem maturity, vendor stability, and price. It is editorial analysis, not a benchmark or investment advice.
What Changed
- ChatGPT retook first place in the Aug. 17 Implicator LLM Meter at 88, with Claude falling to 84 after a single edition in the lead.
- OpenAI's enterprise business passed its ChatGPT-led consumer operation in revenue at an Aug. 14 shareholder meeting, on an unaudited $40 billion annualized run rate that was 20% above June.
- Anthropic's text watermark applies worldwide to Claude models launched since Aug. 2, with no setting, plan tier or API parameter to disable it, including on the platform API and through AWS, Google Cloud and Microsoft Foundry.
- Grok gained five points to 32 after Grok 4.6 shipped Aug. 12 with same-day API access, but stayed ninth because it has no system card or red-team report.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
OpenAI's enterprise crossover
The Aug. 14 shareholder meeting put OpenAI's enterprise business above its ChatGPT-led consumer operation in revenue, about two quarters ahead of its end-2026 forecast. OpenAI's annualized revenue run rate reached $40 billion in July, 20% above June, while its business customer count rose 32% that month.
Those figures are unaudited. ChatGPT's 88 on Aug. 17 restores its Aug. 5 score, before a Black Hat security disclosure cost it three points.
Claude's watermark rollout
The watermark applies worldwide to Claude models launched since Aug. 2, including premium enterprise access through the platform API, AWS, Google Cloud and Microsoft Foundry. No setting, plan tier or API parameter disables it. Because the scorecard weights enterprise buyers' own risk decisions, every Claude-processed deliverable carries a vendor-detectable marker buyers cannot contract out of.
The mark uses a statistical bias in word choice and can survive when Claude lightly edits a person's own writing, depending on passage length and how much Claude changes. Anthropic says detection weakens as Claude-generated text falls because fewer words carry the watermark.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
Four subscribers to Max, which starts at $100 a month, had canceled by Aug. 17.
Anthropic says it has seen no measurable increase in cancellations since the August announcement. It had 300,000 business customers as of September 2025 and maintains that detection shows Claude processed text, not that Claude wrote it.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Grok gains five points
Grok gained five points to 32 on Aug. 17 but stayed ninth. Grok 4.6 shipped Aug. 12 with same-day access through the xAI API, Cursor, OpenRouter, Vercel and Cloudflare, breaking the production-pipeline constraint that had held its score down. It has no system card or red-team report.
Gemini retook third from Mistral, 82 to 81, one week after losing it.
Anthropic plans a watermark-detection API, but had not published one as of Aug. 17.
Frequently Asked Questions
What is the Implicator LLM Meter?
A weekly editorial scorecard that rates 10 models from 0 to 100 from the perspective of an enterprise buyer, on compliance and security, professional-task quality, reliability and service commitments, ecosystem maturity, vendor stability, and price. It is editorial analysis, not a benchmark and not investment advice.
Why did ChatGPT retake first place?
OpenAI's enterprise business passed its consumer operation in revenue about two quarters ahead of the company's end-2026 forecast, on a $40 billion annualized run rate that was 20% above June, with business customers up 32% that month. Those figures are unaudited. The score of 88 restores what ChatGPT held on Aug. 5, before a Black Hat security disclosure cost it three points.
Why did Claude's score fall?
Anthropic's text watermark applies worldwide to Claude models launched since Aug. 2, and no setting, plan tier or API parameter disables it, including premium enterprise access. Because the scorecard weights an enterprise buyer's own risk decisions, every Claude-processed deliverable carries a vendor-detectable marker the buyer cannot contract out of.
Can the watermark survive editing?
The mark uses a statistical bias in word choice and can survive when Claude lightly edits a person's own writing, depending on passage length and how much Claude changes. Anthropic says detection weakens as the amount of Claude-generated text falls, because fewer words carry the watermark.
How many Claude subscribers have canceled?
Four subscribers to Max, which starts at $100 a month, had canceled by Aug. 17. Anthropic says it has seen no measurable increase in cancellations since the August announcement, and it had 300,000 business customers as of September 2025.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
