iCreatAI News
Skip to content
Trending storyWatching

Logical Intelligence Sudoku test: Kona 96.2%, mainstream LLMs only 2%

1 reports1 sources2 days agoUpdated

Understand this story

AI overview

Logical Intelligence's public Sudoku demonstration claims that its energy model Kona solved 96.2% of difficult puzzles in about a week, with an average time of 313 milliseconds. In the same test, GPT-5.2, Claude Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and DeepSeek V3.2 achieved a combined solve rate of only 2%, with GPT-5.2 performing best at just 6.9%. The above figures all come from Logical Intelligence's demonstration and are not an independent evaluation.

AI overview based on source reports · Updated 2 days ago

Coverage timeline

Follow the coverage to explore different perspectives.

Nov 18
  1. Hacker News AI
    What Sudoku Reveals About AI Architecture and the Future of Our Economy

    A demo has Kona solving hard Sudoku puzzles in real time against GPT-5.2, Claude Opus 4.5 and Sonnet 4.5, Gemini 3 Pro and DeepSeek V3.2, with Kona solving 96.2% of puzzles at an average of 313 milliseconds while the frontier LLMs together reached only a 2% solve rate, taking up to 90 seconds before an incorrect answer or timeout. The post attributes this to LLM architecture: autoregressive models generate tokens sequentially and cannot easily revise earlier decisions, whereas Kona produces a complete candidate grid, scores it against all constraints simultaneously via a scalar energy value, and revises using gradient information in continuous latent space. It reports that running Kona on nearly 13,000 solved puzzles cost about $4 in GPU time and roughly 1.1 hours of compute, while LLM API calls for the same demo at a 98% failure rate cost around $11,000.

Activity for this story

Current comparable activity
3
Peak comparable activity
10Nov 18 22:00
Change in the past 24 hours
-51%

Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.