Intelligence Index v4.1 at a glance
Artificial Analysis upgraded three agentic evaluations, removed IFBench, and began reporting cost, time, and tokens per task.
What changed with Intelligence Index v4.1
Artificial Analysis released v4.1 of its Intelligence Index on June 15, 2026. It moved to Terminal-Bench 2.1, τ³-Bench Banking, and GDPval-AA v2, dropped IFBench because it no longer separated frontier models, and shifted weight toward agentic tasks.
The update also added per-task cost, time, and token figures. At publication, the top available model was Claude Opus 4.8 (max) at 56, ahead of GPT-5.5 (xhigh) at 55. Reported cost per task ranged from $0.04 for DeepSeek V4 Pro (max) to $1.78 for Opus 4.8 (max).
Intelligence Index v4.1 in numbers
Values are copied from the linked source on the date shown. They are not IA Compa measurements.
- Top available model at launch (Claude Opus 4.8, max)as of · Source for Top available model at launch (Claude Opus 4.8, max): artificialanalysis.ai (opens in a new tab)
- 56 (v4.1)
Why Intelligence Index v4.1 matters for businesses
Cost per task is a more useful number for agencies than price per token, because it reflects how much a model actually spends to finish a job. That makes it easier to estimate the running cost of an automation before building it.
Sources for this story
- Artificial Analysis Intelligence Index v4.1 (opens in a new tab)Official
Artificial Analysis · artificialanalysis.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #benchmarks
- #artificial-analysis
- #intelligence-index
- #agents
