Puzzle portfolio ladder
30 generated puzzles in five families: logic grids, knapsacks, document facts, specifications and bug hunts.
Run Aug 29 – Oct 3, 2026, 30 items.
Method
Six items from each of five generators whose answers are fixed by construction: zebra logic grids, knapsack optimisation, fact lookup in a long document, checking an implementation against its specification, and finding a planted bug. Difficulty rises within each family, and the bank is frozen.
Run at each provider's default reasoning effort, like the ladder above.
Results
| # | Model | Accuracy | Items |
|---|---|---|---|
| 1 | GPT-5.2 | 96.7% | 30 |
| 2 | GPT-5.5 | 96.7% | 30 |
| 3 | Ling 3.1 Flash | 96.3% | 27 |
| 4 | GLM-5.3 (Mistral API) | 92.9% | 28 |
| 5 | Kimi K2.6 | 90.0% | 30 |
| 6 | Claude Opus 4.8 | 90.0% | 30 |
| 7 | Qwen 3.6 Plus | 86.7% | 30 |
| 8 | DeepSeek V4 Pro | 86.7% | 30 |
| 9 | GPT-OSS-120B | 83.3% | 30 |
| 10 | Nex N2.5 Pro (free route) | 82.1% | 28 |
| 11 | Nemotron 3 Ultra (free route) | 77.8% | 27 |
| 12 | Gemma 4 31B (IT) (free route) | 73.3% | 30 |
| 13 | Nemotron 3 Super (free route) | 66.7% | 27 |
| 14 | Ling-3.0-flash-VL (free route) | 63.3% | 30 |
| 15 | GPT-5.4 | 60.0% | 30 |
| 16 | Space Bunny Alpha (stealth release) | 60.0% | 30 |
| 17 | Gemma 4 26B A4B (IT) (free route) | 53.3% | 30 |
| 18 | Laguna XS 2.1 (free route) | 44.8% | 29 |
| 19 | Kimi K2 0905 | 43.3% | 30 |
- GPT-5.4 solves 29 of 30 when it is asked for high reasoning effort.
- Not ranked, too few answered items: Dots3-Note (Preview) (26 of 30), Nex N2.5 Mini (25 of 30), Ling-3.0-flash-fin (24 of 30), Ling-3.0-flash-sante (24 of 30), Qwen 3.8 27B (24 of 30), Laguna S 2.1 (24 of 30), Apodex 1.1 Mini (24 of 30), North Mini Code (21 of 30), GLM-5.2 (21 of 30), Nemotron 3.5 Lightning (19 of 30), Nemotron 3 Nano Omni 30B-A3B (Reasoning) (19 of 30), LFM2.5-2.6B (12 of 30), Inkling Small (0 of 30), MiniMax-M3 (0 of 30), MiniMax-M2.7 (0 of 30).
In the main leaderboard
These results are collected for the main leaderboard alongside external benchmarks. New scores enter the rating through the regular audited data updates. Comparisons with that rating therefore share some of its inputs.
Interactive version: theaggregate.ai/our-benchmarks/puzzle-portfolio · How It Works · Data refreshed daily, snapshot 2026-10-11.