# SemiAnalysis releases GDPx benchmark for frontier AI models

Digest AI · Research · published 2026-11-02T01:51:42Z

Canonical: https://digestai.news/story/semianalysis-releases-gdpx-benchmark-for-frontier-ai-models

## Summary

SemiAnalysis has launched GDPx, a new benchmark designed to measure the general agentic capabilities of frontier AI models. The benchmark focuses on high-value domains such as chip design, cybersecurity, biology, finance, legal, and software engineering, arguing that these sectors will drive the first trillion dollars of AI revenue. Unlike previous benchmarks like OpenAI's GDPval, which relied on static, non-interactive tasks, GDPx uses realistic, rich environments provided by 14 specialized data partners. For example, finance tasks involve navigating thousands of anonymized files in a virtual data room, while hardware tasks run on live AWS FPGA instances.

The results show that even the best-performing models score only 12% on the full benchmark, indicating significant room for improvement. OpenAI and Anthropic models generally lead the rankings, with Opus 5.5 and GPT-6 Astra performing strongly, though their scores are sometimes lowered by safety classifiers. Gemini 4 Carbon is noted as a strong competitor, particularly in biology and software engineering. The authors highlight that safety classifiers significantly impact scores in domains like cyber and alignment, and that current models struggle heavily with complex finance and legal tasks, where many scored zero.

SemiAnalysis argues that existing benchmarks are inadequate for evaluating today's agents because they lack realistic work environments and often contain unfair grading criteria. By partnering with domain experts, GDPx aims to provide a more accurate picture of where frontier models can actually be deployed at high ROI. The benchmark is available for public exploration, and the authors suggest that tracking progress in these specific high-impact domains is more useful than broad GDP-based metrics.

## Key points

- SemiAnalysis launched GDPx, a benchmark covering 11 high-value domains like chip design and finance.
- The best model scored 12% on GDPx, with OpenAI and Anthropic models leading the rankings.
- GDPx uses realistic environments from 14 partners, unlike static benchmarks like OpenAI's GDPval.

## Why it matters

GDPx provides a more realistic assessment of AI capabilities in high-value industries, helping businesses and researchers identify where frontier models can actually deliver ROI versus where they still fail.

## Sources

1. [GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains](https://newsletter.semianalysis.com/p/gdpx-measuring-frontier-model-capabilities) (SemiAnalysis, 2026-11-02)

Part of the developing story: [AI Model Wars From Replit to Claude Opus](https://digestai.news/thread/replit-ends-partnership-with-anthropic-cites-cost-and-performance) (6 stories)

## Cite

Digest AI, "SemiAnalysis releases GDPx benchmark for frontier AI models", 2 November 2026, https://digestai.news/story/semianalysis-releases-gdpx-benchmark-for-frontier-ai-models

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/semianalysis-releases-gdpx-benchmark-for-frontier-ai-models.json