Skip to content
SemiAnalysis Launches GDPx Benchmark, Opus 5.5 Tops With 12% Score
Depth
12%Opus 5.5 achieved the highest score on the benchmark with a result of 12% · reported
1 outlet · newest source 492h ago
AI hardware

SemiAnalysis Launches GDPx Benchmark, Opus 5.5 Tops With 12% Score

SemiAnalysis introduced a new agentic capability benchmark covering eleven domains, finding Opus 5.5 leads with a 12% score across tasks like chip design and cybersecurity.

Key points

  • SemiAnalysis launched GDPx, a benchmark measuring agentic capabilities across eleven distinct domains.
  • Opus 5.5 achieved the highest score on the benchmark with a result of 12%.
  • The test covers areas including chip design, cybersecurity, RSI, bio, finance, legal, quant trading, software engineering, materials science, and alignment.

SemiAnalysis introduced a new measurement tool for artificial intelligence agents, according to a report by Techmeme. The firm launched GDPx, a benchmark designed to evaluate agentic capabilities across a wide range of specialized fields.SemiAnalysis created a new test called GDPx to see how well AI agents can do complex jobs.SemiAnalysis released GDPx, a multi-domain agentic benchmark where Opus 5.5 achieved a leading 12% score. This release provides a structured way to compare how different models perform in complex, multi-step tasks that require autonomous decision-making and execution.

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

What changed

The publication of GDPx marks a shift toward evaluating AI systems on their ability to act rather than just generate text. SemiAnalysis defined the scope of the benchmark to include eleven specific domains. These areas span technical and professional fields, including chip design, cybersecurity, robotics simulation and integration, biology, finance, law, quantitative trading, software engineering, materials science, and alignment.

How it works

The benchmark measures performance in each of the eleven domains to create a comprehensive score. Opus 5.5 emerged as the top performer in this initial evaluation. The model secured a score of 12 percent, which represents the highest result among the systems tested. This metric suggests a baseline for agentic capability, though the absolute number indicates significant room for improvement in autonomous task completion.

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

What the evidence means

The 12 percent score for Opus 5.5 highlights the current limitations of AI agents in specialized domains. While Opus 5.5 leads the pack, a double-digit score implies that most tasks remain unsolved or require substantial human oversight. The breadth of the benchmark, covering fields from legal analysis to hardware design, underscores the diversity of challenges agents face.

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

Cost and market position

SemiAnalysis positioned itself as a provider of deep technical analysis with this launch. By creating a proprietary benchmark, the firm influences how the market evaluates vendor performance. Opus 5.5’s lead suggests that Anthropic’s model is currently viewed as the most capable agent in this specific test environment. However, the narrow margin and low absolute score mean market leadership in agentic tasks is still fluid.

Evidence: reportedSemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)

aifleet-techmeme

Why it matters

  • Product teams can use the eleven domains to identify which AI agents are ready for specific technical workflows.
  • The low 12% score indicates that current models are not yet reliable for fully autonomous execution in complex fields.
  • SemiAnalysis’s benchmark provides a new metric for comparing vendor capabilities beyond standard language generation tests.
  • Teams working in chip design or cybersecurity have a specific benchmark to track progress in their domains.

Topics: AI benchmarks, SemiAnalysis, Opus 5.5, agentic AI, machine learning

Updated

Then / now / next

Then

AI benchmarks primarily focused on language generation and reasoning tasks.

Now

SemiAnalysis has introduced GDPx to measure agentic capabilities across eleven specialized domains.

Next

Watch for updated scores as other vendors release models optimized for these agentic tasks.

Story timeline

  1. Then AI benchmarks primarily focused on language generation and reasoning tasks.
  2. 02 Nov 2026 SemiAnalysis launches GDPx, an agentic capability benchmark measuring 11 domains, including chip design, cybersecurity, and RSI; Opus 5.5 tops with a 12% score (SemiAnalysis)
  3. Now SemiAnalysis has introduced GDPx to measure agentic capabilities across eleven specialized domains.
  4. Next Watch for updated scores as other vendors release models optimized for these agentic tasks.

The strongest doubt

Counterpoint from the sources

The absolute score of 12% suggests that current AI agents are far from fully autonomous in these complex domains.

For readers

Researchers

  • SemiAnalysis defines agentic capability across eleven specific technical domains.

Investors

  • Opus 5.5 leads the new benchmark, signaling current market leadership in agentic tasks.

Developers

  • The benchmark includes software engineering and chip design, relevant for technical workflows.

Ask the article

Local keyword search of this article; no generated answers.

Did this article give you the evidence you needed?

No vote saved on this device.
Your choice stays on this device. No public tally.

Terms in this story

GDPx

A benchmark launched by SemiAnalysis to measure agentic AI capabilities.

Agentic capability

The ability of an AI system to perform autonomous tasks and make decisions.

Opus 5.5

An AI model that achieved the highest score in the GDPx benchmark.

Questions readers ask

What is the GDPx benchmark?

GDPx is an agentic capability benchmark launched by SemiAnalysis that measures performance across eleven domains, including chip design and cybersecurity.

Which model scored highest on GDPx?

Opus 5.5 topped the benchmark with a score of 12%.

What domains does the benchmark cover?

It covers RSI, hardware design, cyber, bio, finance, legal, quant trading, chip design, software engineering, materials science, and alignment.

Connected stories

This storyHugging Face tests flag UMT5 attention bug on CPUHugging Face tests flag UMT5 attAnalysts link 1.5 GW of smuggled Nvidia chips to Huang’s departureAnalysts link 1.5 GW of smuggledThinking Machines Inkling 2 Small ties Gemini, releases weights next weekThinking Machines Inkling 2 SmalOpenAI confirms resolution of GPT Image 2 API edit request failuresOpenAI confirms resolution of GP

Reading receipt

  • SemiAnalysis launched GDPx, a benchmark measuring agentic capabilities across eleven distinct domains.
  • Opus 5.5 achieved the highest score on the benchmark with a result of 12%.
  • The test covers areas including chip design, cybersecurity, RSI, bio, finance, legal, quant trading, software engineering, materials science, and alignment.

Still open: How do other major models compare to Opus 5.5 on this specific benchmark?

Still open: What is the methodology for calculating the 12% score?

Time spent: 0:00 · Depth: Full

Sources

  1. Techmeme

Related stories

Background reading