All Cryptomat articles tagged #benchmarks.

Vals AI released its Recursive Self-Improvement Index on Tuesday, scoring how well frontier models can build successor systems and reigniting the AI…

SpaceXAI's Grok 4.5 topped GPT-5.6-SOL on reasoning benchmarks, per a Crypto Briefing report Sunday, reopening the AI-token trade after a summer of chop.
Daily at 6 AM UTC. Serious analysis. No hype. No ads.