R2ABench (Mini-SWE-Agent) - Syntax Validity: leaderboard

Requirements-to-architecture generation on R2ABench (2026): from a structured requirements specification of 68 educational and GitHub projects, the model driving the Mini-SWE-Agent framework writes a PlantUML architecture view, compared with a human-validated reference. Share of views that render and parse into an architecture graph, from 0 to 1; higher is better.

Metric: Syntax-valid rate (0-1): share of generated views that render and parse into an architecture graph (L0), over the 68 R2ABench projects (17 educational and 51 GitHub repositories), generating a PlantUML architecture view from the structured requirements specification with the Mini-SWE-Agent framework under default settings and greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4.60.9118#179
2GPT-50.8382#185
3Qwen 3 Coder 480B A35B0.7794#716
4DeepSeek V3.20.7647#425

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=r2abench-mini-swe-agent-syntax-validity · How It Works · Data refreshed daily, snapshot 2026-11-22.