# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-193307
**Date:** Apr 02, 2026
**Category:** analysis
**Question ID:** ANALYSIS-013

---

## Question

A mobile app shows: DAU 100K (up 50%), WAU 200K (up 20%), MAU 500K (up 10%), D1 retention 40%, D7 retention 15%, D30 retention 5%, Session length 8min (down 20%), Sessions/day 2.1 (up 30%). The PM says 'we're growing fast.' (1) What story does this data actually tell? (2) What's concerning about the DAU/MAU ratio? (3) Why might session length decrease as sessions increase? (4) What one metric would you focus on and why?

---

## Winner

**Claude Sonnet 4.6** (openrouter)
- Winner Score: 9.43
- Matrix Average: 8.91
- Total Judgments: 90

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | Claude Sonnet 4.6 | openrouter | 9.43 | 9 |
| 2 | Claude Opus 4.6 | openrouter | 9.18 | 9 |
| 3 | Grok 4.20 | openrouter | 9.13 | 9 |
| 4 | GPT-5.4 | openrouter | 9.07 | 9 |
| 5 | GPT-OSS-120B | OpenAI | 9.04 | 9 |
| 6 | MiniMax M2.5 | openrouter | 8.99 | 9 |
| 7 | Gemini 3 Flash Preview | Google | 8.93 | 9 |
| 8 | DeepSeek V4 | openrouter | 8.87 | 9 |
| 9 | MiMo-V2-Flash | Xiaomi | 8.70 | 9 |
| 10 | Gemini 3.1 Pro | openrouter | 7.72 | 9 |

---

## 10×10 Judgment Matrix

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

| Judge ↓ / Resp → | GPT-OSS-120B | Gemini 3.1 Pro | Claude Opus | GPT-5.4 | DeepSeek V4 | MiMo-V2-Flash | Claude Sonnet | Grok 4.20 | Gemini 3 | MiniMax M2.5 |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-OSS-120B | — | 6.7 | 8.7 | 8.4 | 8.4 | 8.7 | 8.8 | 8.7 | 8.3 | 8.4 |
| Gemini 3.1 Pro | 8.5 | — | 8.8 | 9.8 | 9.8 | 6.3 | 10.0 | 9.6 | 9.8 | 8.8 |
| Claude Opus | 9.4 | 7.5 | — | 9.2 | 8.0 | 9.2 | 9.8 | 9.2 | 9.2 | 9.2 |
| GPT-5.4 | 8.0 | 6.3 | 8.2 | — | 8.8 | 9.0 | 9.2 | 8.8 | 8.6 | 8.8 |
| DeepSeek V4 | 9.0 | 8.6 | 10.0 | 9.0 | — | 9.0 | 9.6 | 9.0 | 9.0 | 8.8 |
| MiMo-V2-Flash | 9.2 | 8.6 | 9.6 | 8.8 | 9.2 | — | 9.6 | 9.3 | 9.3 | 9.3 |
| Claude Sonnet | 9.6 | 8.1 | 9.6 | 8.8 | 8.6 | 8.8 | — | 8.8 | 8.8 | 9.0 |
| Grok 4.20 | 8.8 | 8.1 | 8.8 | 8.8 | 8.8 | 8.8 | 9.0 | — | 8.6 | 8.8 |
| Gemini 3 | 10.0 | 8.8 | 10.0 | 9.8 | 9.8 | 9.8 | 10.0 | 9.8 | — | 9.8 |
| MiniMax M2.5 | 9.0 | 7.0 | 9.0 | 9.0 | 8.4 | 8.8 | 9.0 | 9.0 | 8.8 | — |

---

## 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: ANALYSIS-013. 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-20260402-193307/results
Full dataset: https://app.themultivac.com/dashboard/export