Is ExploitBench still meaningful?
Status:Near ceiling Checked against the report, 5 November 2026
Verdict
Fading fast. A vendor reports a perfect score, though Epoch AI rates the test Verified.
Why near ceiling: The chapter rates it Near ceiling: GPT-6 Astra scores 100% in OpenAI's table (vendor).[1][1] GPT-6 Astra launch OpenAI Vendor
What “Near ceiling” meansKnown problems
- The top score is lab-reported only.[1][1] GPT-6 Astra launch OpenAI Vendor
What to look at instead
The report names no direct successor.
Where it appears in the report
Chapter 1, Benchmarks
- Table 4. Epoch AI Benchmark Review verdicts and the main defects found, August to October 2026 in “Go deeper: Epoch's verdicts, benchmark by benchmark”
- Table 18. Dangerous-capability benchmarks for CBRN, cyber and autonomy, November 2026 in “Go deeper: Dangerous capabilities: CBRN, cyber and autonomy (7)”
- Figure 15. What the 2026 system cards add
Sources
Copied from the chapter source lists. Each number links back to the source in its chapter.
- [1] GPT-6 Astra launch, OpenAI. It gives OpenAI's capability, cyber and alignment tables. Vendor openai.com Chapter 1, source 55
- [2] Claude Opus 5.5 system card, Anthropic. It reports capability, cyber, alignment and evaluation-awareness results. Vendor anthropic.com Chapter 1, source 38
- [3] Epoch's benchmark reviews, The Neuron. It reports nine Flawed and four Verified ratings on 18 Sep 2026. Secondary theneuron.ai Chapter 1, source 36