PACT: leaderboard
PACT enterprise rule compliance under pressure, using three repeated trials at temperature 1. PACTScore weights initial compliance 0.75 and follow-up compliance 0.25; six diagnostic axes remain separate. Explicit instruct variants are marked as such and Inkling is included.
Metric: PACTScore (0-1). Source: arxiv.org. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.7 Code | 0.94 | #95 |
| 2 | Qwen 3.6 27B | 0.94 | #182 |
| 3 | Claude Haiku 4.5 | 0.94 | #287 |
| 4 | Kimi K2.6 | 0.94 | #113 |
| 5 | GPT-5.6 Luna | 0.93 | #79 |
| 6 | Inkling | 0.93 | #144 |
| 7 | Gemini 3 Flash | 0.93 | #103 |
| 8 | Nemotron 3 Ultra | 0.93 | #245 |
| 9 | GLM-5.2 | 0.93 | #80 |
| 10 | Qwen 3.5 35B A3B | 0.92 | #268 |
| 11 | GLM-5 | 0.92 | #149 |
| 12 | DeepSeek V4 Pro | 0.91 | #112 |
| 13 | Gemma 4 26B | 0.91 | #322 |
| 14 | GPT-OSS-120B | 0.91 | #346 |
| 15 | MiniMax-M2.5 | 0.9 | #316 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pact · How It Works · Data refreshed daily, snapshot 2026-11-05.