Open-Weights Models Near Frontier in PINNACLE Benchmark, Undercutting Pause Calls

Signal65's updated PINNACLE agent benchmark shows open-weight models closing in on the frontier—Qwen3.8 outperforming Claude Opus 5 on error rate and four enterprise dimensions within 15 days—undercutting lab calls to slow frontier development.

2026-09-15 ~ 2026-09-16 · 3 related posts

• Signal65 tests show open-weight models closing on the frontier, undercutting pace calls — ryanshrout · 2026-09-15
• PINNACLE Agentic Benchmark: Qwen3.8 Makes 15% Fewer Errors Than Claude Opus 5 — ryanshrout · 2026-09-15
• Open-weight Qwen3.8 beats Claude Opus 5 on 4 of 5 enterprise personas just 15 days after launch — ryanshrout · 2026-09-16