Opus 5 Ranks First With 61 Points, Costs 26% Less Than Fable 5

Opus 5 leads with a 61-point index and $2.03 task cost, undercutting Fable 5 by 26%. It tops knowledge work tests but faces a 50% hallucination rate.

Woofun AI data shows that Opus 5 achieved a comprehensive intelligence index score of 61, narrowly surpassing Fable 5’s 60 points, while GPT-5.6 Sol and Kimi K3 scored 59 and 57 respectively. The model’s average single-task cost stands at $2.03, representing a 26% reduction compared to Fable 5’s $2.75. Opus 5 topped the GDPval-AA v2 and AA-Briefcase evaluations and tied for first in the Programming Agent index when integrated with Claude Code, achieving an 89% score on Terminal-Bench v2.1.

The system provides five reasoning intensity levels, where token output varies by approximately eight times between low and maximum settings, correlating with a 407 Elo difference in GDPval-AA v2 scores. Despite these gains, Opus 5 trails Fable 5 in factual knowledge accuracy. Its hallucination rate in the AA-Omniscience test rose to 50%, a 14 percentage point increase from Opus 4.8, and its cost-effectiveness at lower reasoning levels remains slightly below the GPT-5.6 series.