Microsoft's CLIO Hits 75.2% on Agent's Last Exam
Microsoft published new benchmark results for Microsoft Discovery, its platform for organizations doing Frontier research and development, in a post on the Azure Blog written by Aseem Datar, Corporate Vice President of Product Innovation.
Transcript
Microsoft says Discovery Engine with CLIO outscored the other agent systems tested.
On Agent's Last Exam, a professional task benchmark, CLIO scored seventy five point two percent in physical sciences.
CLIO runs independent reasoning paths that explore a problem, share learning, and resolve into one evidence backed result, bringing in a human expert when needed.
Microsoft says the system already supported work that discovered a new organic redox flow battery. It points to chip design and drug discovery.
One caveat. These are Microsoft's own scores on its own blog, with no independent audit, against unnamed systems Microsoft chose. Microsoft sells this product.
Inboxsmith builds AI phone and email agents for small business. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- On Agent's Last Exam, a demanding evaluation of long-running, tool-using professional tasks, Microsoft Discovery Engine with CLIO (Cognitive Loop via In-Situ Optimization) achieved higher scores than the other agentic harnesses evaluated across three scientific domains: 61.6% in health and medicine, 75.2% in physical sciences, and 64.6% in life sciences.(Microsoft's official announcement, "Beyond the benchmark: How an adaptive approach drives scientific discovery", published September 8, 2026)
- CLIO enables independent reasoning paths to explore a problem, compare and share learning, and resolve the strongest trajectory into a single evidence-backed result. The system can determine when to keep exploring, change strategy, use a different model, or bring a domain expert into the loop.(Microsoft's official announcement, "Beyond the benchmark: How an adaptive approach drives scientific discovery", published September 8, 2026)
- Discovery Engine with CLIO has already supported work that discovered a novel organic redox flow battery. The same approach has potential across design simulation (like for silicon chips), formulation and process optimization (for example in manufacturing and CPG), materials and molecular discovery (which can drive sustainability and drug discovery), and lab automation, areas where organizations need to shorten research cycles, without sacrificing rigor or traceability.(Microsoft's official announcement, "Beyond the benchmark: How an adaptive approach drives scientific discovery", published September 8, 2026)
- HONESTY CAVEAT: This report rests entirely on Microsoft's own announcement, published September 8, 2026 on Microsoft's corporate Azure blog and written by a Microsoft corporate vice president. The benchmark scores on Agent's Last Exam are Microsoft's own reported results; the post cites no independent audit and no third party verification of those numbers. The comparison is stated only against "the other agentic harnesses evaluated", a comparison set Microsoft itself chose and does not name in the post, so the margin over specific competing systems cannot be checked. Microsoft also sells Microsoft Discovery, the commercial platform this result promotes, so the announcement is both a research claim and product marketing. No outside reporting or corroborating source was available at the time of writing.(Microsoft's official announcement, "Beyond the benchmark: How an adaptive approach drives scientific discovery", published September 8, 2026)