> ## Content Index
> Fetch the complete content index at: https://www.implicator.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# LLM Meter — Week of May 31
- URL: https://www.implicator.ai/llm-meter-week-of-may-31/
- Published: 2026-06-01T03:18:34.000Z
- Updated: 2026-06-01T03:18:34.000Z
- Author: Marcus Schuler
- Tags: #llm-meter

—CLAUDE---  
score: 91  
trend: up  
change: +5  
\+ Opus 4.8 (May 28) hits 69.2% on SWE-bench Pro, about ten points ahead of GPT-5.5, retaking the coding lead enterprises benchmark against  
\+ $65B Series H at a $965B valuation passes OpenAI (\~$852B) to make Anthropic the most valuable AI lab, with run-rate revenue tripling to \~$47B  
\+ Opus 4.8 shipped day-one on Amazon Bedrock, Google Vertex AI, and Microsoft Foundry, the three clouds regulated buyers already procure through  
\+ Claude Code "dynamic workflows" (hundreds of parallel subagents for codebase migrations) lands on Enterprise, Team, and Max plans  
\- June 15 Agent SDK metering, opaque plan-limit denominators, and a brief May 28 billing blip keep the procurement-trust drag alive  
\---GEMINI---  
score: 88  
trend: down  
change: -2  
\+ The former Vertex AI consolidates into the Gemini Enterprise Agent Platform, unifying training, deployment, and agent orchestration in one stack  
\+ I/O adopters (Salesforce Agentforce, Databricks, Ramp, Xero) plus Spark and Deep Research Max keep Google's distribution procurement-grade  
\- Flagship Gemini 3.5 Pro slips to June, leaving Flash to carry the lineup the week Anthropic ships a new top-of-field Opus  
\- A quiet post-I/O week cedes both the coding lead and the most-valuable-lab title to Anthropic, dropping Gemini out of first  
\- DeepMind's Pentagon-work letter and classified-network questions remain unresolved  
\---CHATGPT---  
score: 84  
trend: down  
change: -1  
\+ ChatGPT Enterprise added Skills governance, tighter upload scanning, and expanded Compliance Logs Platform support  
\+ Codex gains Windows Computer Use plus a GitHub Enterprise Server template for customer-hosted repos, extending into on-prem workflows  
\- GPT-5.5 now trails Opus 4.8 by about ten points on SWE-bench Pro, ceding the enterprise coding-agent quality lead  
\- Anthropic's $965B raise eclipses OpenAI's \~$852B valuation, ending its run as the most valuable AI lab  
\- About $14B in projected 2026 losses against the $600B compute commitment keeps the financial overhang in place  
\---MISTRAL---  
score: 73  
trend: up  
change: +1  
\+ Airbus partnership spans commercial aircraft, helicopters, defense, and space, a marquee European industrial-and-defense reference account  
\+ New "Vibe" platform (rebranded Le Chat) pushes Mistral from model vendor into enterprise workflow software for email, documents, and code  
\+ Mistral for Industrial Engineering folds Emmi AI physics simulation into a stack for aerospace, automotive, and semiconductor buyers  
\+ €4B France/Sweden data-center buildout and a stated plan to explore custom chips reinforce the EU-sovereign supply story  
\- Benchmarks still trail Opus 4.8, GPT-5.5, and Gemini 3.5 Flash, and Mistral remains outside the Pentagon classified-network roster  
\---GROK---  
score: 30  
trend: down  
change: -1  
\+ Grok V9-Medium (1.5T parameters, trained on Cursor developer data) finished training May 25, with a coding-focused release targeted for mid-June  
\+ Quality Mode for the Grok Imagine API adds higher-realism generation for enterprise creative teams  
\- No federal, compliance, or enterprise-procurement progress while rivals shipped flagships and a record raise  
\- The SpaceXAI structure and $1.25B/month Anthropic compute deal cast xAI more as a GPU landlord to rivals than a Grok enterprise vendor  
\- "Reckless financials" framing persists after the SpaceX IPO disclosed heavy cash burn against \~$500M ARR  
\---DEEPSEEK---  
score: 17  
trend: up  
change: +1  
\+ Reported funding round nears a \~$44-45B valuation (Big Fund in talks to lead, Tencent and Alibaba participating), \~4x the figure floated two weeks earlier  
\+ Permanent 75% V4-Pro price cut to \~$0.44 per million input tokens holds the global cost-leadership floor, under a tenth of GPT-5.5  
\- The round is still "in talks," not closed, and Big Fund state backing deepens the US-procurement compliance problem  
\- V4-Pro quality still trails Western flagships (\~9th globally), a gap that just widened against Opus 4.8  
\- US government-device bans and the broader compliance perimeter remain fully in force