Day-one recall check
The 34-question recall test every new free model gets on the day it appears.
Run Aug 31 – Oct 3, 2026, 34 items.
Method
A short version of the book recall bank: 34 questions about 12 public-domain books in three tiers, famous, mid and obscure. It is the first step of the battery we run on every new free model the day it appears, so it is the one recall board where free models sit next to the flagships.
Graded by matching the key, with no second-pass judge. Saying “unknown” counts as wrong; an empty reply does not count at all. Reasoning effort was set to high where the route accepts it.
Results
| # | Model | Accuracy | Items |
|---|---|---|---|
| 1 | Claude Fable 5 | 61.8% | 34 |
| 2 | Claude Opus 5 | 58.8% | 34 |
| 3 | Claude Fable 5.1 | 55.9% | 34 |
| 4 | GPT-5.5 | 47.1% | 34 |
| 5 | Claude Opus 4.8 | 41.2% | 34 |
| 6 | GPT-5.2 | 38.2% | 34 |
| 7 | GPT-5.4 | 38.2% | 34 |
| 8 | Kimi K2.6 | 32.4% | 34 |
| 9 | DeepSeek V4 Pro | 32.4% | 34 |
| 10 | Nex N2.5 Mini (free route) | 32.4% | 34 |
| 11 | Nemotron 3 Ultra (free route) | 32.3% | 31 |
| 12 | GPT-5.4 Mini | 29.4% | 34 |
| 13 | Qwen 3.6 35B A3B | 29.4% | 34 |
| 14 | Nex N2.5 Pro (free route) | 29.4% | 34 |
| 15 | Nemotron 3 Super (free route) | 26.5% | 34 |
| 16 | Kimi K2 0905 | 23.5% | 34 |
| 17 | Qwen 3.6 Plus | 23.5% | 34 |
| 18 | Ling-3.0-flash-fin (free route) | 23.5% | 34 |
| 19 | GPT-OSS-120B | 20.6% | 34 |
| 20 | Nemotron 3.5 Lightning (free route) | 17.6% | 34 |
| 21 | Ling-3.0-flash-sante (free route) | 17.6% | 34 |
| 22 | Ling-3.0-flash-VL (free route) | 17.6% | 34 |
| 23 | Apodex 1.1 Mini (free route) | 17.6% | 34 |
| 24 | Gemma 4 26B A4B (IT) (free route) | 14.7% | 34 |
| 25 | Gemma 4 31B (IT) (free route) | 14.7% | 34 |
| 26 | Space Bunny Alpha (stealth release) | 14.7% | 34 |
| 27 | Ling 3.1 Flash | 14.7% | 34 |
| 28 | Dots3-Note (Preview) (free route) | 11.8% | 34 |
| 29 | Claude Haiku 4.5 | 8.8% | 34 |
| 30 | Laguna XS 2.1 (free route) | 8.8% | 34 |
| 31 | North Mini Code (free route) | 8.8% | 34 |
| 32 | Laguna S 2.1 (free route) | 6.2% | 32 |
| 33 | LFM2.5-2.6B (free route) | 2.9% | 34 |
- One question is 3 points here, and 34 questions could not separate Claude Fable 5 from Opus 5; the book recall bank above is the precise version.
- Not ranked, too many empty replies: Nemotron 3 Nano Omni 30B-A3B (Reasoning) (28 of 34), GLM-5.2 (26 of 34), GLM-5.3 (25 of 34), Qwen 3.8 27B (12 of 34), Inkling Small (0 of 34), Inkling (0 of 34), MiniMax-M3 (0 of 34), MiniMax-M2.7 (0 of 34).
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/recall-check · How It Works · Data refreshed daily, snapshot 2026-10-11.