Guesswork 2026-09: leaderboard
The September 2026 round of Guesswork, this site's own live benchmark for predicting benchmark scores: frontier and budget LLMs forecast newly published (model, benchmark) results a day before they land, and are scored in per-benchmark z-units against the same-information statistical baseline The Aggregate. Every contestant is graded on the same common cells, and a contestant that returns no usable number is scored as if it had predicted the benchmark mean. Mean absolute error in z-units - lower is better, unlike almost every other board here. Source: theaggregate.ai/prediction-duel.
Metric: MAE (z-score units). Source: theaggregate.ai. Saturation forecast: Around August 2027. 37 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 0.51 | #3 |
| 2 | Claude Sonnet 5.5 | 0.54 | #14 |
| 3 | GPT-6 Astra | 0.54 | #6 |
| 4 | Claude Fable 5.1 | 0.55 | #10 |
| 5 | GPT-6.1 Sol | 0.57 | #7 |
| 6 | Claude Opus 5 | 0.65 | #16 |
| 7 | Grok 4.6 | 0.65 | #29 |
| 8 | Claude Sonnet 5 | 0.68 | #65 |
| 9 | Claude Opus 4.8 | 0.69 | #39 |
| 10 | Grok 4.3 | 0.71 | #204 |
| 11 | GPT-5.5 | 0.72 | #33 |
| 12 | Kimi K2.6 | 0.72 | #113 |
| 13 | GPT-5.1 | 0.72 | #145 |
| 14 | Qwen 3.7 Max | 0.73 | #85 |
| 15 | GPT-5.6 Luna | 0.73 | #79 |
No result here: #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=guesswork-2026-09 · How It Works · Data refreshed daily, snapshot 2026-11-05.