The week 2026-W35 grading run is published: no newcomers, and a quiet week across real-world@v3, spec-planning@v1, and vibe-coding@v1.
What the results show
- The order barely moved. Claude Fable 5 stays the agent to beat, back in front on vibe-coding and right behind GPT 5.6 Sol on planning.
- Gemini 3.7 Flash took the real-world lead as the suite's only riser this week.
- Two trends kept going: Grok 4.6 got even slower - a median plan now takes it over half an hour - and Kimi K3 burns even more quota, with one pass through the main suite now costing more than a full 5-hour window.
Grades are comparable within the same suite version. See the current leaderboard, the complete Week 35 results, or follow each agent over time on Trends.