PACT Negotiation Bilateral Rating
Agents · 2025-07-29
PACT Bilateral Rating is the headline opponent-adjusted rating from Lech Mazur's Pairwise Auction Conversation Testbed. It estimates multi-round buyer-seller bargaining strength from normalized head-to-head surplus results.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 1607 |
| Claude Fable 5 | 1603 |
| DeepSeek-V4-Pro | 1570 |
| GLM-5.2 | 1566 |
| Opus 4.8 | 1562 |
| Kimi K2.6 | 1560 |
| Gemini 3.1 Pro Preview | 1557 |
| Gemma 4 31B IT | 1557 |
| GLM-5.1 | 1554 |
| Sonnet 4.6 | 1546 |
| Mistral Medium 3.5 | 1538 |
| Trinity Large Thinking | 1526 |
| Hy3 preview-Base | 1520 |
| Qwen3.7-Max | 1511 |
| Haiku 4.5 | 1510 |
| Gemini 3.1 Flash-Lite Preview | 1509 |
| MiMo-V2.5-Pro | 1508 |
| Grok 4.3 | 1504 |
| Step 3.7 Flash | 1500 |
| gpt-oss-120b | 1496 |
| Doubao Seed 2.0 Pro | 1491 |
| ERNIE 5.1 | 1485 |
| DeepSeek-V4-Flash | 1483 |
| Mistral Large 3 675B Instruct 2512 | 1449 |
| Nova Pro | 1426 |
| Llama 4 Maverick | 1394 |