> ## 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 24
- URL: https://www.implicator.ai/llm-meter-week-of-may-24/
- Published: 2026-05-26T04:02:24.000Z
- Updated: 2026-05-26T04:02:24.000Z
- Author: Marcus Schuler
- Tags: #llm-meter

\---GEMINI---  
score: 90  
trend: up  
change: +2  
\+ Gemini 3.5 Flash scores 55 on Artificial Analysis, within 2 points of Claude Opus 4.7 and 5 of GPT-5.5 at roughly a third of the per-token cost  
\+ Named enterprise adopters at launch (Salesforce Agentforce, Databricks, Ramp, Xero) turn I/O into procurement-grade distribution  
\+ I/O cleared last week's post-launch downside risk; the feared "lands behind GPT-5.5" framing did not materialize  
\+ Antigravity agent-platform upgrade plus Google Spark extend the agentic stack enterprise buyers can standardize on  
\- Flagship Gemini 3.5 Pro slips to next month and the DeepMind Pentagon-work letter stays unresolved  
\---CLAUDE---  
score: 86  
trend: down  
change: -1  
\+ Claude Opus 4.7 remains the reference flagship rivals benchmark against, holding the top of the quality field  
\- Gemini 3.5 Flash now lands within two points of Opus 4.7 at a third of the price, pressuring Claude's premium positioning  
\- June 15 Agent SDK metering plus unpublished plan-limit denominators keep the procurement-trust drag live  
\- No new enterprise win this week after the PwC/SAP/Gates run, ceding the cycle to Google and OpenAI  
\- Pentagon IL6/IL7 classified-network exclusion still unresolved  
\---CHATGPT---  
score: 85  
trend: up  
change: +1  
\+ OpenAI-Dell deal brings Codex to hybrid and on-prem environments via the Dell AI Data Platform, a real regulated-buyer unlock  
\+ Gartner named OpenAI a Leader in enterprise coding agents, a procurement-cycle credential  
\+ New $234M Singapore Applied AI Lab, OpenAI's first outside the US, deepens sovereign-enterprise reach  
\+ GPT-5.5 now fully rolled out across Plus, Business, and Enterprise tiers  
\- Roughly $14B in projected 2026 losses and the $600B compute commitment keep the financial overhang in place  
\---MISTRAL---  
score: 72  
trend: down  
change: -1  
\+ Acquired Vienna's Emmi AI to add physics-aware industrial modeling to the enterprise stack  
\+ EU-sovereign procurement option intact via the Paris/Sweden buildout and AI Act compliance dossier  
\- A quiet week with only a tuck-in acquisition while every top-tier rival shipped a major enterprise signal  
\- Benchmark standing trails Opus 4.7, GPT-5.5, and now Gemini 3.5 Flash on price/performance  
\- Remains outside the Pentagon classified-network vendor roster  
\---GROK---  
score: 31  
trend: up  
change: +1  
\+ Heavy product cadence: Grok Skills (May 18), third-party connectors (May 22), and the Grok Build coding agent landed in quick succession  
\+ Grok V9 Medium completed training (May 25), reportedly 1.5T parameters, signaling roadmap momentum  
\- SpaceX's IPO filing disclosed xAI burned $6.4B last year, with another round of layoffs and a Grok-team restructure  
\- New features are developer/consumer-facing; no federal, compliance, or enterprise-procurement progress  
\- "Chatbot no one uses" narrative persisted into this week, capping any adoption credit  
\---DEEPSEEK---  
score: 16  
trend: up  
change: +1  
\+ Permanently cut V4-Pro API price 75% to \~$0.44 per million input tokens, under a tenth of GPT-5.5, reinforcing cost leadership  
\+ First external funding round ($45-50B, Tencent/Alibaba/Big Fund) continues, easing the undercapitalization argument  
\- V4-Pro ranks 9th globally at 63.87% accuracy; the quality gap to Western flagships persists  
\- The aggressive cut also reads as margin pressure as open-weight rivals close the gap  
\- US government-device bans and the broader compliance perimeter remain fully in force