TetrisBench
Gemini Flash reaches 66% win rate on Tetris against Opus
Details
- External ID
- 46769752
- Source
- HN
- Company
- —
- Product
- TetrisBench
- Website domain
- tetrisbench.com
- Launched
- Jan. 26, 2026
- Cohort
- —
- Upvotes
- 111
- Upvotes percentile
- 0.9077733860342556
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:25 p.m.
- Updated at
- Sept. 7, 2026, 9:25 p.m.
Enrichment
- Theme
- multimodal AI models and agents
- Vertical
- Horizontal
- Function
- Observability & eval
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Hobby / open-source project
- Normalized one-liner
- benchmark comparing llm performance on tetris
- Manually corrected
- False
Could you build this?
Yes The project is a benchmarking harness and static web dashboard that feeds Tetris state to various LLM APIs and records win/loss ratios.
Discussion
20 comments analyzed.
Competitors mentioned: Stockfish (chess engine), Reinforcement learning with genetic algorithms, OpenCV with algorithmic parsing, Hatetris (adversarial piece selection variant)
Concerns raised: Token economics and context accumulation costs grow expensive quickly, LLM latency too slow for real-time Tetris gameplay (seconds vs milliseconds), Over-optimizes short-term line clears at expense of long-term stability, Not a meaningful benchmark for LLM capabilities, Tetris Holdings aggressive trademark enforcement
Feature requests: Email optional/not required to play, Support for more randomizer variants (7-bag, custom randomizers)
Competitors
Other products that read as similar to this one — 34 launches clear the similarity bar, closest 8 shown.
Attention rank: #5 of 35 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 48 days after the earliest competitor.
- Redactle LLM Leaderboard · hn · 2026-09-03 · 5 upvotes · similarity 0.45
- tetris · github · 2026-09-28 · 17 upvotes · similarity 0.44
- jev-column-race · github · 2026-09-17 · 20 upvotes · similarity 0.41
- Fly By · hn · 2026-09-01 · 84 upvotes · similarity 0.39
- Gemini free tier is all you need · hn · 2026-04-30 · 8 upvotes · similarity 0.38
- We beat Gemini Embedding 2 by training only 16M params (open weights) · hn · 2026-07-10 · 7 upvotes · similarity 0.37
- Google Gemini 3.8 Flash and Cyber · ph · 2026-09-04 · 237 upvotes · similarity 0.37
- Gemini 3.1 Flash Live · ph · 2026-03-27 · 320 upvotes · similarity 0.36
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a observability & eval tool for Media & entertainment yet.