# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-153453
**Date:** Jan 28, 2026
**Category:** reasoning
**Question ID:** REASON-003

---

## Question

Estimate how many piano tuners there are in Chicago. Show your reasoning step by step, including:
1. All assumptions you make
2. How you derived each number
3. Sensitivity analysis (what if your assumptions are wrong?)
4. Final estimate with confidence interval

---

## Winner

**GPT-5.4** (openrouter)
- Winner Score: 9.07
- Matrix Average: 8.10
- Total Judgments: 80

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | GPT-5.4 | openrouter | 9.07 | 9 |
| 2 | GPT-OSS-120B | OpenAI | 9.05 | 9 |
| 3 | Claude Opus 4.6 | openrouter | 8.90 | 9 |
| 4 | MiMo-V2-Flash | Xiaomi | 8.74 | 9 |
| 5 | Grok 4.20 | openrouter | 8.74 | 9 |
| 6 | Claude Sonnet 4.6 | openrouter | 8.70 | 9 |
| 7 | Gemini 2.5 Flash | openrouter | 8.63 | 9 |
| 8 | DeepSeek V4 | openrouter | 8.42 | 9 |
| 9 | Gemini 3.1 Pro | openrouter | 2.68 | 8 |

---

## 10×10 Judgment Matrix

Rows = Judge, Columns = Respondent. Self-judgments excluded (—).

| Judge ↓ / Resp → | Gemini 3.1 Pro | DeepSeek V4 | Claude Opus | GPT-5.4 | Grok 4.20 | Claude Sonnet | MiMo-V2-Flash | GPT-OSS-120B | Gemini 2.5 | MiniMax M2.5 |
|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3.1 Pro | — | 8.8 | 10.0 | 9.8 | 8.8 | 8.3 | 10.0 | 9.2 | 8.8 | · |
| DeepSeek V4 | 6.0 | — | 9.1 | 8.7 | 8.7 | 8.7 | 9.2 | 8.7 | 8.7 | · |
| Claude Opus | 1.6 | 7.8 | — | 8.9 | 8.9 | 8.9 | 8.6 | 9.2 | 8.9 | · |
| GPT-5.4 | 0.7 | 7.8 | 8.2 | — | 8.6 | 8.6 | 7.9 | 8.2 | 8.4 | · |
| Grok 4.20 | 4.0 | 8.4 | 8.7 | 8.7 | — | 8.7 | 8.7 | 9.2 | 8.7 | · |
| Claude Sonnet | 1.4 | 8.2 | 9.2 | 9.2 | 8.6 | — | 8.9 | 9.2 | 8.6 | · |
| MiMo-V2-Flash | 2.2 | 9.0 | 9.4 | 9.4 | 9.4 | 9.4 | — | 9.4 | 9.0 | · |
| GPT-OSS-120B | 3.3 | 8.2 | 8.4 | 8.4 | 8.4 | 8.4 | 8.2 | — | 8.3 | · |
| Gemini 2.5 | · | 9.2 | 8.8 | 9.2 | 9.0 | 9.1 | 8.8 | 10.0 | — | · |
| MiniMax M2.5 | 2.1 | 8.6 | 8.4 | 9.4 | 8.3 | 8.4 | 8.4 | 8.6 | 8.4 | — |

---

## Methodology

- **10×10 Blind Peer Matrix:** All models answer the same question, then all models judge all responses.
- **5 Criteria:** Correctness, completeness, clarity, depth, usefulness (each scored 1–10).
- **Self-judgments excluded:** Models do not judge their own responses.
- **Weighted Score:** Composite of all 5 criteria.

---

## Citation

The Multivac (2026). Blind Peer Evaluation: REASON-003. app.themultivac.com

## License

Open data. Free to use, share, and build upon. Please cite The Multivac when using this data.

Download raw JSON: https://app.themultivac.com/api/evaluations/EVAL-20260402-153453/results
Full dataset: https://app.themultivac.com/dashboard/export