Terra, Gemini Pro, Grok, Qwen Max and Kimi refreshed
140 successful model/prompt builds published in autonomous sessions with browser tools. See each run for its output, transcript, runtime and cost.
GPT-5.6 Terra, Gemini 3.1 Pro Preview, and Kimi K3 each completed all 35 prompts. Grok 4.7 completed 18 and Qwen 3.8 Max completed 17. The remaining 35 attempts failed when OpenRouter credits ran out; those failures are recorded on the affected runs.
Top builds in this update
OUTPUT gpt-5.6-terra Top-down alien shooter
Generated pi 0.87.1 · browser iteration · 63.6s · $0.0610 · 5.0k out
Full run + transcript →
OUTPUT gemini-3.1-pro-preview Top-down alien shooter
Generated pi 0.87.1 · browser iteration · 84.8s · $0.0982 · 6.7k out
Full run + transcript →
OUTPUT grok-4.7 Top-down alien shooter
Generated pi 0.87.1 · browser iteration · 493.5s · $0.7738 · 27k out
Full run + transcript →
OUTPUT qwen3.8-max-0902 Top-down alien shooter
Generated pi 0.87.1 · browser iteration · 724.7s · $0.5490 · 24k out
Full run + transcript →