CritPt v2 launches with all 70 physics problems rewritten
10/23/2026 — 10/23, 04:22·1 sources (all)·4 reports (all)
Key takeaways
- Artificial Analysis has released CritPt v2, in which the author team and physics experts reviewed and rewrote all 70 research-level physics…
- models may now answer with code execution tools enabled. Artificial Analysis notes v2 scores are higher than v1 and not directly comparable,…
Background & analysis
On October 23, 2026, Artificial Analysis announced on X that CritPt v2 had gone live. According to the post, the CritPt authors, working with physics expert collaborators and responding to feedback from the research community, reviewed and revised all 70 research-level physics challenges. Models can now attempt the questions with access to a code-execution tool, and the full results are available on the CritPt project page.
A second post published at the same time addressed scoring. Artificial Analysis said overall CritPt v2 scores are higher than those on v1 and that the two versions are not directly comparable. GPT-6 Astra (max) scores 83% on v2, up from 32% on v1. In its testing, Artificial Analysis attributed most of the increase to the revised challenges, with code-execution tool access contributing a smaller share.
The public information currently stops there: v2 is live with all 70 questions revised, GPT-6 Astra (max) has moved from 32% to 83%, and the organization stresses that scores across the two versions should not be compared directly.
AI-generated from 2 reports · updated 9 hours ago
Latest turnArtificial Analysis reports that GPT-6 Astra (max) scores 83% on CritPt v2, up from 32% on v1, though the two versions are not directly comparable. In its testing, most of the increase came from the revised challenges, with code-execution tool access contributing a smaller share.
Related AI tools
- Heat index
- 78
- All sources
- 1
- All reports
- 4
- First seen
- 9 hours ago
Reports on this story headlines open the original
Artificial Analysis reports that GPT-6 Astra (max) scores 83% on CritPt v2, up from 32% on v1, though the two versions are not directly comparable. In its testing, most of the increase came from the revised challenges, with code-execution tool access contributing a smaller share.
X：Artificial Analysis (@artificialanlys)First-partyAI score 43Artificial Analysis has launched CritPt v2, with the CritPt authors and physics experts reviewing all 70 research-level physics challenges. Models now attempt them with access to a code-execution tool, updating how the benchmark measures model performance on research-grade physics.
X：Artificial Analysis (@artificialanlys)First-partyAI score 51Artificial Analysis reports that GPT-6 Astra (max) scores 83% on CritPt v2, up from 32% on v1, but warns the two versions are not directly comparable. In its testing, most of the gain comes from the revised challenges, with code-execution tool access adding a smaller share.
X：Artificial Analysis (@artificialanlys)First-partyAI score 49CritPt v2 is live on Artificial Analysis. After feedback from the research community, the benchmark's authors and physics collaborators revised all 70 research-level physics challenges, and models now attempt them with a code-execution tool. Full results are on the CritPt project site.
X：Artificial Analysis (@artificialanlys)First-partyAI score 49
Other stories people are talking about
- Heat index 487METR finds flaws in OpenAI's agent monitoring system11 sources
- Heat index 356SurgeOpenAI opens Decisions API, built on GPT-6 Luna4 sources
- Heat index 213OpenAI brings GPT-6 Astra and Codex to Google Cloud Model Garden6 sources
- Heat index 49Artificial Analysis Benchmarks 14 Models on Long-Context Medical Records1 sources
- Heat index 1181PHP ships emergency patch for two critical AI-found flaws22 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion