Perspective AI

Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on private datasets The Artificial Analysis Intelligence Index is our composite measure of model capabilities across tasks reflecting real-world usage today. v5 is our biggest upgrade since v4 launched in January. With v5, 60% of the Index is now based on evaluations with private test sets or held-out answers. Alongside Intelligence Index v5, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale. New to the Intelligence Index: ➤ FrontierCode 2.0: Cognition's newly-launched private benchmark of real-world coding tasks, graded on whether a repository maintainer would merge the agent's change ➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment Upgraded: ➤ HLE-Diamond (replacing Humanity's Last Exam): a refined 1,000-question version of Humanity's Last Exam, split evenly between reasoning and knowledge questions ➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised ➤ CritPt v2: a private, 40-question hard subset of CritPt v2, revised by the CritPt authors and run with a code-execution tool Staying in the Index: AA-Briefcase, AutomationBench-AA, GDP.pdf and Terminal-Bench 4.0 Removed from the Index: GDPval-AA v2.1, SciCode, and AA-LCR v1.1 Key results: ➤ Claude Opus 5.5 (max) leads at 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55): Anthropic and OpenAI hold the top five models. Gemini 4 Argon (high) is next at 50 ➤ The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not yet released) is the top open weights model at 46. Among released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) lead at 32 - 26 points behind the leader and below Claude Haiku 5.5 (max, 36) ➤ Cost per task varies by more than 13x among the leading models: of the 10 highest-scoring models shown, GPT-6.1 Terra (max) is the cheapest at $0.96 per Intelligence Index task and Claude Fable 5.1 (max) the most expensive at $13.04

Devin Fusion with Claude Opus 5.5 (max) and a GPT-6 Luna (high) sidekick scores 68 on the Coding Agent Index, 2 points above Claude Code with Opus 5.5 (max) at 31% lower cost per task @cognition's Devin Fusion runs a frontier lead model alongside a lower-cost sidekick model. The two work in parallel with their own context and tools, with the lead keeping planning, delegation and final review. We benchmarked Devin Fusion CLI with Claude Opus 5.5 (max) + GPT-6 Luna (high). It scores 68.1 on the Artificial Analysis Coding Agent Index, closely matched at the top with Claude Code with Claude Sonnet 5.5 (max) at 68.4 and Muse Code with Muse Wonder Full (max) at 68.0. Against Claude Code with Claude Opus 5.5 (max) at 66.0, it costs $9.01 vs. $13.04 per task and averages 55 vs. 65 minutes per task. Across the three benchmarks, the Fusion configuration scores higher on DeepSWE v1.1 (72.6% vs. 68.4%) and SWE-Atlas-QnA (69.6% vs. 66.4%) and slightly lower on Terminal-Bench 4.0 (62.1% vs. 63.1%). Its results reflect both models working together, including their combined API cost. Costs are pay-per-token API costs, not subscription prices.

Today the EF and OA announce zkAPI—a new means for private AI Picture this: deposit ether into a vault, sign a zk proof, get a fixed amount of private inference. Advances like these help us move past middlemen's requirements that we link our data and identities. Read more below

THIS ANTHROPIC IPO IS THE MOST DANGEROUS DEAL I'VE SEEN IN MY 45-YEAR CAREER And its own IPO filing tells you why: The prospectus warns that its AI models could resist being shut down and could hide or manipulate information. If you run security at a big bank, that's the last thing you're letting anywhere near your customers' money. The surveys back that up too: 69% of IT and security leaders say security worries are slowing down their AI agent rollouts, and only 1 in 5 American businesses use AI in any part of their work. And the revenue Anthropic does have leans on a VERY short list. Nearly 25% of last year's sales came from just 2 customers, and plenty of its big customers aren't locked into long-term contracts. Now look at the bills: Anthropic has signed up for $518 billion of computing over the next decade, and 80% of it gets paid whether the customers show up or not. The filing says so itself: "If our actual spend falls short, we must pay Google the difference." What could possibly go wrong? Every dollar of that $518 billion sits in somebody else's revenue forecast. Wall Street's tech analysts have the sector's cash flow doubling to $2.4 trillion by 2028, and the analysts who cover the customers are forecasting a much smaller pile of cash to pay for it. Wayne Gretzky's father famously taught him "to skate to where the puck is going, not where it's been." For 3 years the puck was chips and data centers, and the people selling them got rich. Now the puck is heading to the customer, and the customer is SCARED. At $2 trillion you're paying for NARRATIVE DOMINANCE, a story where every company on earth runs on AI. When the customers don't show up on schedule, most of that $518 billion still comes due, and whoever holds the stock eats the difference. We skated to where this puck was going a long time ago: In January we showed you that most CFOs couldn't point to any measurable return on their AI spending. Those CFOs sign the checks this whole thing depends on, and 8 months later their security teams are still standing in the doorway. In May I told you SpaceX's record IPO would be forced into the index funds on its 15th trading day, and that your 401k would be the exit liquidity. On July 7 it joined the Nasdaq-100, and the funds tracking it had to buy an estimated $4.3 billion of stock. Anthropic is next in line. AND THIS IS MUCH MORE DANGEROUS. $2 trillion IPO, record spending, and no customers. If they can't get money from investors it's OVER for the whole AI boom. Own businesses whose customers are paying them today, like energy, and let somebody else pay $2 trillion for customers that haven't shown up yet. Don't be the exit liquidity.