---
title: "Agents · tool hallucination signal: model rankings and source evidence · Undominated.ai"
canonical: https://undominated.ai/rankings/arena-agent-tool-hallucination/
description: "Agents · tool hallucination signal ranked by IPS score. Inspect exact model variants, source dates, measurement limits and downloadable evidence."
---

# Agents · tool hallucination signal: model rankings and source evidence · Undominated.ai

> Agents · tool hallucination signal ranked by IPS score. Inspect exact model variants, source dates, measurement limits and downloadable evidence.

[All rankings](/rankings/) / Coding and agents

# Agents · tool hallucination signal

Agent Arena signal scores from the official dataset. Each signal is a separate measurement; it is not a success percentage.

**58 entries** · IPS score · Higher is better

Source publication: 2026-11-04. Retrieved 2026-11-05.

Source: [Arena (lmarena-ai)](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset) · [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/).

Ranks use confidence bounds. Several models can share a rank; the rows within a shared rank are displayed by score.

Agents · tool hallucination signal · IPS score

| Significance rank | Model / evaluated variant | IPS score | 95% confidence interval | Observations |
| --- | --- | --- | --- | --- |
| 1 | GLM-5.3-Flash zai | 0.008903 | 0.008081 – 0.009725 | 11,871,656 |
| 1 | Grok 4.6 (xHigh) xai | 0.008903 | 0.008081 – 0.009725 | 3,557,471 |
| 1 | GPT-6 Luna (Low) openai | 0.008903 | 0.008081 – 0.009725 | 2,732,091 |
| 1 | GLM-5.3 (Max) zai | 0.008903 | 0.008081 – 0.009725 | 10,241,529 |
| 1 | GPT-6.1 Sol (Max) openai | 0.008903 | 0.008081 – 0.009725 | 2,555,443 |
| 1 | GPT-6 Astra (Max) openai | 0.008903 | 0.008081 – 0.009725 | 1,194,324 |
| 1 | GPT-6.1 Terra (Max) openai | 0.008903 | 0.008081 – 0.009725 | 6,051,267 |
| 1 | Kimi K3.1 Pro (Max) moonshot | 0.008903 | 0.008081 – 0.009725 | 4,948,298 |
| 1 | DeepSeek-V4-Pro (High, 20260813) deepseek | 0.008903 | 0.008081 – 0.009725 | 812,009 |
| 1 | GPT-5.6 Sol (xHigh) openai | 0.008903 | 0.008081 – 0.009725 | 6,106,979 |
| 1 | GPT-6 Luna (Medium) openai | 0.008903 | 0.008081 – 0.009725 | 7,544,280 |
| 1 | Kimi K3 (Max) moonshot | 0.008902 | 0.00808 – 0.009724 | 14,767,204 |
| 1 | GPT-6 Sol (Max) openai | 0.008901 | 0.008079 – 0.009723 | 2,813,701 |
| 1 | Grok 4.7 (xHigh) xai | 0.008887 | 0.008064 – 0.009709 | 3,300,736 |
| 1 | GPT-6 Luna (Max) openai | 0.008878 | 0.008055 – 0.0097 | 36,929,317 |
| 1 | GLM-5.2 (Max) zai | 0.008868 | 0.008043 – 0.009693 | 8,374,554 |
| 1 | GPT-5.6 Terra (xHigh) openai | 0.008754 | 0.007905 – 0.009604 | 3,291,847 |
| 1 | Muse Spark 1.1 meta | 0.008512 | 0.007617 – 0.009408 | 7,669,418 |
| 1 | Muse Spark 1.3 (Max) meta | 0.008438 | 0.00755 – 0.009325 | 7,804,971 |
| 1 | Claude Opus 5 (Max) anthropic | 0.00843 | 0.007507 – 0.009352 | 4,255,614 |
| 1 | Claude Fable 5.1 (Max) anthropic | 0.007679 | 0.006376 – 0.008983 | 2,223,941 |
| 1 | Claude Sonnet 5.5 (Max) anthropic | 0.007632 | 0.00628 – 0.008985 | 1,040,676 |
| 1 | Muse Spark 1.2 (xHigh) meta | 0.007242 | 0.005166 – 0.009317 | 3,315,530 |
| 1 | Claude Opus 5.5 (Max) anthropic | 0.006573 | 0.004703 – 0.008443 | 2,043,959 |
| 1 | Claude Fable 5 (High) anthropic | 0.006125 | 0.004138 – 0.008112 | 3,140,962 |
| 21 | Gemini 3.7 Flash (High) google | 0.005845 | 0.004396 – 0.007294 | 5,594,205 |
| 21 | Gemini 4 Carbon (High) google | 0.005128 | 0.003127 – 0.007128 | 6,732,336 |
| 21 | Claude Sonnet 5 (High) anthropic | 0.005119 | 0.003673 – 0.006564 | 4,781,638 |
| 23 | Gemini 4 Argon (High) google | 0.004166 | 0.00269 – 0.005643 | 5,809,460 |
| 23 | Gemini 3.8 Flash (High) google | 0.004115 | 0.002526 – 0.005704 | 7,341,646 |
| 25 | DeepSeek-V4.1-Flash-20261017 (Max) deepseek | 0.003483 | 0.002426 – 0.00454 | 17,216,312 |
| 24 | MiniMax-M3.1-Flash minimax | 0.002606 | 0.000389 – 0.004822 | 2,191,528 |
| 29 | Hy4 preview tencent | 0.000898 | -0.001114 – 0.00291 | 10,128,437 |
| 30 | Qwen3.8-27B alibaba | 0.000757 | -0.001146 – 0.00266 | 7,754,962 |
| 23 | Gemini 3.1 Pro Preview google | -0.000254 | -0.006393 – 0.005885 | 2,736,062 |
| 32 | MiMo-V2.6-Pro xiaomi | -0.000567 | -0.002963 – 0.001829 | 4,554,607 |
| 1 | Muse Wonder Full (Max) meta | -0.001526 | -0.013961 – 0.01091 | 5,071,624 |
| 1 | Mistral Large 4 mistral | -0.00161 | -0.012066 – 0.008845 | 386,267 |
| 32 | Qwen3.7-Max alibaba | -0.002866 | -0.007023 – 0.001291 | 2,107,566 |
| 28 | Hy4 tencent | -0.003706 | -0.010677 – 0.003265 | 1,314,198 |
| 33 | Step 5 Preview stepfun | -0.004843 | -0.008598 – -0.001088 | 2,427,836 |
| 36 | Qwen3.8-Max alibaba | -0.007335 | -0.009506 – -0.005164 | 7,343,430 |
| 1 | Muse Wonder (Max) meta | -0.007339 | -0.023436 – 0.008759 | 1,745,681 |
| 38 | Qwen3.8-Flash-Next alibaba | -0.008963 | -0.010838 – -0.007088 | 15,320,065 |
| 36 | MiMo-V2.5-Pro xiaomi | -0.009511 | -0.013193 – -0.00583 | 2,319,225 |
| 35 | Beam 1.0 501B A23B reflection | -0.010394 | -0.019257 – -0.00153 | 555,707 |
| 19 | MiniMax-M3 minimax | -0.01354 | -0.03468 – 0.0076 | 3,464,873 |
| 38 | Claude Haiku 5.5 (High) anthropic | -0.014679 | -0.021543 – -0.007815 | 3,371,105 |
| 21 | Nemotron 3 Ultra nvidia | -0.017226 | -0.041399 – 0.006948 | 209,335 |
| 37 | Spark-X2.6 synlan | -0.019553 | -0.032155 – -0.00695 | 856,437 |
| 36 | Inkling thinky | -0.020014 | -0.034229 – -0.0058 | 1,645,936 |
| 39 | Qwen3.7-Plus alibaba | -0.021772 | -0.034357 – -0.009186 | 1,177,145 |
| 43 | Inkling-Small thinky | -0.02793 | -0.043249 – -0.012611 | 499,930 |
| 46 | Solar Pro 4 upstage | -0.037426 | -0.055087 – -0.019766 | 488,500 |
| 46 | MiMo-V2.6-Flash xiaomi | -0.041881 | -0.063616 – -0.020147 | 3,121,990 |
| 48 | Mistral Medium 3.5 mistral | -0.045178 | -0.065144 – -0.025212 | 599,909 |
| 48 | Hy3 tencent | -0.046969 | -0.06997 – -0.023969 | 1,149,676 |
| 55 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) nvidia | -0.072922 | -0.086881 – -0.058963 | 1,336,586 |

## What this ranking can tell you

 - The population is the models in this source board, including variants and models without an available API price.
- Ranks use the published confidence bounds: one plus the number of models whose lower bound exceeds this model’s upper bound. Shared ranks are not proof of equal performance.

## Check the source

The original evidence is retained byte for byte. The downloadable board includes the source URLs, revision and SHA-256 hashes.

[Download this board and provenance →](/data/rankings/arena-agent-tool-hallucination.json)

 Source files and licences - [metadata · a7a159f557a1](/data/rankings/evidence/a7a159f557a1e594a5d41218fe8d53c63c63daa39b655e51d6fc5b482a17d059.json) · [upstream source ↗](https://huggingface.co/api/datasets/lmarena-ai/leaderboard-dataset)
- [licence · 88a46d472f35](/data/rankings/evidence/88a46d472f35274fb6b7a852c3d16c8ee8007d51b37f290161537bec26895ee4.txt) · [upstream source ↗](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/raw/8b6deb69ace01ee8dcda9e46eba7a5e8a808ec53/README.md)
- [rows · bb1b9d40eeca](/data/rankings/evidence/bb1b9d40eeca30c1ab54a3b90717a21e2daa4d34bb356054851b3bb7f3390fe6.parquet) · [upstream source ↗](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/resolve/8b6deb69ace01ee8dcda9e46eba7a5e8a808ec53/agent_tool_hallucination/latest-00000-of-00001.parquet)

[Explore other model types →](/rankings/)[Read the benchmark guide →](/benchmarks/)

## Continue your investigation

 - [Understand the measurements](/benchmarks/)
- [Compare task performance](/best/)
- [Compare embedding models](/embeddings/)
- [Inspect ranking rules](/methodology/)