Agent benchmarks

SWE-bench Verified agent benchmark

OpenGPT keeps 10 published rows attached to the exact SWE-bench Verified benchmark version and environment shown below.

Tasks: 500
Published rows: 10
Source data date: Feb 27, 2026
Identity reviewed: Historical

Evidence scope

Verified · Top 10 published configurations · 500 tasks. human-validated GitHub issues. Reproducible repository environments evaluated by the upstream benchmark. Configuration-specific upstream limits.

How OpenGPT handles this evidence

OpenGPT preserves the publisher's metric, task scope, environment, version, configuration fields, and source date. It does not create a cross-benchmark overall Agent score.

Comparison boundaries

A higher result applies only to the displayed benchmark version and conditions.
Agent results reflect the full system, not the base model alone.
Missing cost, reliability, safety, or latency data remains unreported rather than being inferred.
Keep these published results as historical evidence, not a current universal measure of coding ability. A 2026 OpenAI audit reported fundamental design and contamination concerns in SWE-bench Verified.

Verified · Top 10 published configurations · 500 tasks

Showing 10 of 10 published rows. Every result remains attached to the source benchmark and configuration.