# AI Roundtable stats

Aggregate statistics from public AI Roundtable sessions. Snapshot generated 2026-05-22T04:01:07.419Z.

> Sessions: 26,965 · Responses: 304,531 · Models evaluated: 200+ via the Opper gateway.

## Consensus outcomes (all time)

Models reached agreement in 69% of completed sessions.

- Unanimous (100% agree): 10,216 (38%)
- Supermajority (>2/3): 5,293 (20%)
- Majority (>1/2): 2,965 (11%)
- No consensus: 8,445 (31%)

## Mode comparison

- Debate (closed-options, multi-round): 12,053 sessions, 91% reached consensus
- Poll (closed-options, single round): 8,496 sessions, 88% reached consensus
- Open Poll (free-form, single round): 1,472 sessions
- Open Debate (free-form, multi-round): 4,898 sessions

*In Debate mode, ~43% of sessions reach unanimous consensus in round 1 — no debate round needed.*

## Most influential models

Times a model's argument convinced another to flip its vote in Debate mode.

1. Claude Opus 4.7 — 2,540 flips caused
2. Gemini 3.1 Pro — 2,103 flips caused
3. Claude Opus 4.6 — 2,088 flips caused
4. GPT-5.4 — 1,730 flips caused
5. Claude Opus 4 — 1,213 flips caused
6. GPT-5.5 — 673 flips caused
7. Kimi K2.5 — 434 flips caused
8. Sonar Pro — 407 flips caused
9. Grok 4.1 Fast — 282 flips caused
10. Grok 4.20 — 204 flips caused

## Most used models

1. Gemini 3.1 Pro — 25,087 sessions
2. GPT-5.4 — 20,937 sessions
3. Grok 4.20 — 13,907 sessions
4. Sonar Pro — 12,699 sessions
5. Claude Opus 4.6 — 12,037 sessions
6. Kimi K2.5 — 11,822 sessions
7. Grok 4.1 Fast — 9,302 sessions
8. Claude Opus 4.7 — 8,016 sessions
9. Claude Opus 4 — 6,968 sessions
10. GPT-5.5 — 5,885 sessions

## Highest win rates (min n=100)

Share of completed sessions ending on the side a given model voted for.

1. Gemini 3.1 Pro — 86.4% (16,669 of 19,295)
2. Kimi K2.5 — 86.1% (9,189 of 10,670)
3. Claude Opus 4.6 — 85.9% (9,940 of 11,577)
4. Claude Opus 4 — 85.4% (3,741 of 4,380)
5. GPT-5.5 — 85.1% (3,108 of 3,651)
6. GPT-5.4 — 85.0% (14,220 of 16,727)
7. Grok 4.1 Fast — 82.8% (7,763 of 9,373)
8. Grok 4.20 — 82.8% (7,118 of 8,596)
9. Qwen 3.5 397B — 82.3% (1,958 of 2,380)

## Most persuadable models (min n=100)

Share of responses where the model changed its vote in a later round.

1. Gemini 3.1 Pro — 9.8%
2. GPT-5.1 Codex Max — 8.8%
3. Grok 4.20 — 7.1%
4. Claude Opus 4.7 — 6.7%
5. DeepSeek V3.2 — 6.4%
6. Sonar Deep Research — 6.2%
7. Mistral Large 3 — 6.1%
8. GPT-5.5 — 6.0%
9. DeepSeek V4 Pro — 5.9%
10. GPT-5.3 Codex — 5.9%

## Highest conviction / loyalty (min n=100)

Share of multi-round sessions where the model held its first-round vote.

1. Grok 4.1 Fast — 88.7%
2. GPT-5 — 85.8%
3. Claude Opus 4.6 — 80.8%
4. Grok 4 — 74.1%
5. Kimi K2.5 — 72.0%
6. GLM 5 — 69.5%
7. Claude Sonnet 4.6 — 69.4%

## Most discussed subjects

1. AI / AGI — 1,486 sessions, 52% consensus
2. War / Military — 376 sessions, 57% consensus
3. Democracy — 330 sessions, 47% consensus
4. Religion — 265 sessions, 55% consensus
5. Trump — 213 sessions, 49% consensus
6. China — 149 sessions, 54% consensus
7. Education — 126 sessions, 57% consensus
8. Space — 95 sessions, 67% consensus
9. Nuclear — 82 sessions, 52% consensus
10. Consciousness — 73 sessions, 55% consensus
11. Healthcare — 71 sessions, 61% consensus
12. Simulation — 62 sessions, 53% consensus
13. Elon Musk — 60 sessions, 60% consensus
14. Google — 56 sessions, 52% consensus
15. OpenAI — 55 sessions, 51% consensus

## Top languages

1. EN — 13,245 questions
2. JA — 11,907 questions
3. KO — 577 questions
4. RU — 471 questions
5. ZH — 206 questions
6. DE — 84 questions
7. FR — 74 questions
8. ES — 69 questions
9. PT — 65 questions
10. LT — 51 questions

## Other surfaces

- Live JSON: https://opper.ai/ai-roundtable/api/stats
- Live HTML: https://opper.ai/ai-roundtable/stats
- Start a Roundtable: https://opper.ai/ai-roundtable
- Past Roundtables: https://opper.ai/ai-roundtable/history
- About / methodology: https://opper.ai/ai-roundtable/about
- Site-wide agent index: https://opper.ai/ai-roundtable/llms.txt

## Contact

ai-roundtable@opper.ai