BlogLLM Benchmark Scores as Published, Side by Side
SRE-Bench: Results as Published
Every result the catalog holds for SRE-Bench: 3 models from 2 developers, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area CybersecurityMaintainer Vals AI publishes and hosts the benchmark pageIntroduced 2026Data version 2026103001, as of 2026-10-30
SRE-Bench is a reverse-engineering benchmark, not a site-reliability one: it tests whether AI agents can work out what a compiled program does when no source code is given. Agents use a shell sandbox with standard analysis tools across five areas (network protocols, firmware, games, file-format recovery and malware). The benchmark has 262 binaries and 1,572 tasks that are graded deterministically.
SRE-Bench
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-6 AstraOpenAI, released 2026-09-03 | 88% | OpenAI, GPT-6 Astrachecked 2026-10-14 | - |
| GPT-5.6 SolOpenAI, released 2026-07-09 | 55.9% | OpenAI, GPT-6 Astrachecked 2026-10-14 | - |
| Claude Opus 5Anthropic, released 2026-07-24 | 12.5% | OpenAI, GPT-6 Astrachecked 2026-10-14 | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.