# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260207-134919
**Date:** Mar 18, 2026
**Category:** reasoning
**Question ID:** REASON-010

---

## Question

You're a consultant charging $500/hour. A client asks you to find the optimal solution to a complex problem. You estimate:
- A quick solution (1 hour) has 60% chance of being optimal
- More analysis (5 hours) has 90% chance of being optimal
- The optimal solution saves the client $50,000 vs. the suboptimal one

1. How much analysis should you do?
2. How would your answer change if you charged $1000/hour?
3. What if you were doing this for yourself (no billing)?
4. Generalize: derive a formula for optimal thinking time given problem stakes and thinking cost

---

## Winner

**MiMo-V2-Flash** (Xiaomi)
- Winner Score: 9.73
- Matrix Average: 8.66
- Total Judgments: 90

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | MiMo-V2-Flash | Xiaomi | 9.73 | 6 |
| 2 | GPT-OSS-120B | OpenAI | 9.73 | 7 |
| 3 | Grok 3 (Direct) | xAI | 9.61 | 7 |
| 4 | DeepSeek V3.2 | DeepSeek | 9.60 | 7 |
| 5 | Claude Sonnet 4.5 | Anthropic | 9.51 | 7 |
| 6 | Claude Opus 4.5 | Anthropic | 9.43 | 7 |
| 7 | Gemini 3 Flash Preview | Google | 9.39 | 7 |
| 8 | Gemini 2.5 Flash | Google | 9.39 | 6 |
| 9 | OLMo Think | Allen AI | 5.63 | 6 |
| 10 | Gemini 3 Pro Preview | Google | 4.59 | 7 |

---

## 10×10 Judgment Matrix

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

| Judge ↓ / Resp → | MiMo-V2-Flash | Gemini 3 | GPT-OSS-120B | Gemini 3 | Claude Sonnet | DeepSeek V3.2 | Gemini 2.5 | Claude Opus | OLMo Think | Grok 3 |
|---|---|---|---|---|---|---|---|---|---|---|
| MiMo-V2-Flash | — | 2.8 | 10.0 | 9.6 | 9.3 | 9.3 | 9.6 | 9.2 | 8.0 | 9.8 |
| Gemini 3 | 0.0 | — | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| GPT-OSS-120B | 0.0 | 0.0 | — | 8.8 | 9.4 | 0.0 | 0.0 | 8.8 | 0.0 | 9.1 |
| Gemini 3 | 10.0 | 6.2 | 10.0 | — | 10.0 | 10.0 | 10.0 | 10.0 | 0.0 | 10.0 |
| Claude Sonnet | 9.8 | 2.6 | 9.8 | 9.2 | — | 9.0 | 9.3 | 9.2 | 0.5 | 9.8 |
| DeepSeek V3.2 | 9.8 | 4.7 | 9.4 | 9.4 | 9.4 | — | 9.6 | 9.4 | 8.0 | 9.4 |
| Gemini 2.5 | 10.0 | 9.3 | 10.0 | 10.0 | 10.0 | 10.0 | — | 10.0 | 10.0 | 10.0 |
| Claude Opus | 9.2 | 1.6 | 9.2 | 9.2 | 9.2 | 9.7 | 8.4 | — | 0.3 | 9.2 |
| OLMo Think | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 10.0 | 0.0 | 0.0 | — | 0.0 |
| Grok 3 | 9.6 | 4.8 | 9.7 | 9.6 | 9.2 | 9.2 | 9.4 | 9.4 | 7.0 | — |

---

## 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-010. 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-20260207-134919/results
Full dataset: https://app.themultivac.com/dashboard/export