Gemini 3.7 Flash speedruns Pokemon via code execution
DynamicWebPaige · x · 2026-09-01
A benchmark from Google DeepMind shows Gemini 3.7 Flash speedrunning the Kanto region in Pokemon using 1/3 the turns (22k vs 70k). Logs reveal the model literally writes Python code and spins up a sandbox to simulate game actions like pushing boulders before making moves.
Related event: Gemini 3.7 Flash Beats Pokémon Benchmark in Far Fewer Steps by Writing Code(2 posts)→
More from Models
- Users are running 'abliterated' GLM-5.3 models locally without safety guardrails — cephaloform · 2026-09-02
- AI fails silently and accumulates inaccuracies over time, unlike humans — gerardsans · 2026-09-02
- Grok 4.6 leads in biosecurity refusal without compromising research utility — ns123abc · 2026-09-02
- Anthropic investigating elevated errors on Claude for Microsoft 365 (Sep 1) — ClaudeAI-mod-bot · 2026-09-02
- Multi-model pipelines become standard; Gemini 3.7 Flash acts as a low-cost auditor — DynamicWebPaige · 2026-09-01
- Users Report Claude Coding Failures: "I Don't Know What to Do" — doooyle · 2026-09-01