TECCI - Gemini Generated Instruction Set: leaderboard
The generated-instruction set of TECCI (Agrawal et al., 2026): five Gemini-written edit instructions of set types per image across seven categories. Overall human success rate on 1,050 sampled edits, in percent, requiring a mean rating of 4.5 of 5 on all three criteria; higher is better.
Metric: Overall human success rate (%; five trained raters score each edit 1-5 and a criterion succeeds when the mean is at least 4.5 on all three criteria; 1,050 sampled pairs, 30 images per category with all five generated instructions). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nano Banana Pro | 23 | |
| 2 | Grok Imagine Pro | 21.3 | |
| 3 | Nano Banana 2 (Gemini 3.1 Flash Image Preview) | 20.2 | |
| 4 | Seedream 5.0 Lite | 18.2 | |
| 5 | GPT Image 1.5 | 5.9 |
Interactive version: theaggregate.ai/benchmark?slug=tecci-gemini-generated-instruction-set · How It Works · Data refreshed daily, snapshot 2026-11-12.