Next.js evals: Grok 4.7 scores 94% vs 97% for GPT-6 Sol and Opus 5.5

Nova @NovaXCodeOriginally posted 2026.09.23

Open original on X

grok 4.7 is not what you think 💀 next.js evals: > opus 5.5, gpt-6 sol, fable 5.1 hit 97% > grok 4.7 hits 94% 3% less on a test. 2-7x cheaper in reality. same jobs, more output per dollar. that’s the actual win.