Semgrep benchmarked OpenAI's new Astra model against 13 points higher than Luna, scoring 100% on ExploitBench.
Public source
Publisher name
Public post
OpenAI's new Astra model destroyed every other model we've benchmarked -- 13points higher than Luna, previously our winner. It also scored 100% on ExploitBench (previous…
Company
Semgrep
Semgrep is the leader in code security for builders, helping teams catch and fix real security issues before they ship.
- Industry
- Software Development
- Location
- San Francisco, US
- Company size
- 201–500 employees
About Semgrep
Semgrep is the leader in code security for builders. Teams catch, flag, and fix real issues before they ship, powered by security that learns as you build. Built for builders and trusted by security, the platform unifies SAST, SCA, and secrets scanning, embedding protection directly into the development workflow so security begins where code is written and lives where developers work. Semgrep combines deterministic static analysis with AI reasoning to power detection, triage, and remediation. This approach helps teams uncover real vulnerabilities, prioritize reachable risks, and fix issues faster. Customers report up to 80% fewer false positives across Code and Supply Chain, with 95% of findings validated by security reviewers across more than 6 million results. Founded in San Francisco, Semgrep is backed by Menlo Ventures, Felicis Ventures, Lightspeed Venture Partners, Redpoint Ventures, and Sequoia Capital. It is recognized by Gartner in Application Security Testing and trusted by leading organizations, including Snowflake, Dropbox, and Figma. Learn more at semgrep.dev.
See moreLatest activity
Latest activity from Semgrep
6 signals
Presence & Recognition
Semgrep's Director of Global Partnerships, Eric Snyder, attended the AWS Startup Partner Summit this week.
Research & Knowledge
Semgrep published internal evaluations showing that Astra scored a perfect 100% on the ARC-AGI-3 benchmark with 52% fewer actions compared to the human baseline.
Research & Knowledge
Semgrep achieved a 100% SOTA performance on the ARC-AGI-3 benchmark according to François Chollet in 6 months from 1%.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Sonar
Sonar published an evaluation of OpenAI's GPT-5.6 Sol and Terra, highlighting security profile shifts and code writing differences.
Research & Knowledge
Vals AI
Vals AI reports that Astra effectively tied Fable 5.1 at 90% for the #1 spot on the Vibe Code Bench benchmark.
Research & Knowledge
CodeRabbit
CodeRabbit put GPT-6 Astra through code review evaluations to compare its performance against GPT-5.6 Sol and Opus 5.
Research & Knowledge
Endor Labs
Endor Labs AI SAST found 192 real vulnerabilities in a benchmark compared to the next leading tool, with 2.6x more true positives and 63 findings caught.
Research & Knowledge
Tigera