POPA Recon#AI #xAI #Grok #LLM ⚡️
#AI #xAI #Grok #LLM ⚡️ 🤔 xAI drops Grok 4.5 - the "Opus-class" model that costs 4x less to run Musk calls it Opus-level, faster, and token-efficient. Benchmarks are more nuanced - but the economics might not care. On SWE-Bench Pro: 64.7% - behind Opus 4.8 (69.2%) and Fable (80.4%). On Terminal Bench 2.1: 83.3%, nearly tied with GPT 5.5. Takes 1st place on Harvey Legal Agent Benchmark despite its coding-first positioning The real edge: 15,954 output tokens per task vs 67,020 for Opus 4.8 - a 4.2x efficiency gap. API: $2/$6 per 1M tokens. Live in Grok Build, Cursor, and xAI API now ✨ Not a…
The excerpt is quoted from a public Telegram channel, which owns its text.