---
title: "Agents · bash recovery (historical): model rankings and source evidence · Undominated.ai"
canonical: https://undominated.ai/rankings/arena-agent-bash-recovery-steps/
description: "Agents · bash recovery (historical) ranked by IPS score. Inspect exact model variants, source dates, measurement limits and downloadable evidence."
---

# Agents · bash recovery (historical): model rankings and source evidence · Undominated.ai

> Agents · bash recovery (historical) ranked by IPS score. Inspect exact model variants, source dates, measurement limits and downloadable evidence.

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

# Agents · bash recovery (historical)

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

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

Source publication: 2026-10-08. 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 · bash recovery (historical) · IPS score

| Significance rank | Model / evaluated variant | IPS score | 95% confidence interval | Observations |
| --- | --- | --- | --- | --- |
| 1 | Claude Sonnet 5.5 (Max) anthropic | 0.150352 | 0.133322 – 0.167383 | 9,166 |
| 1 | Claude Opus 5 (Max) anthropic | 0.135701 | 0.127779 – 0.143624 | 60,324 |
| 1 | Claude Opus 5.5 (High) anthropic | 0.135534 | 0.127511 – 0.143556 | 23,322 |
| 1 | Claude Opus 5 (High) anthropic | 0.126375 | 0.118669 – 0.134081 | 77,004 |
| 1 | Claude Fable 5.1 (Max) anthropic | 0.116032 | 0.092874 – 0.13919 | 36,016 |
| 5 | Claude Fable 5 (High) anthropic | 0.095976 | 0.083758 – 0.108194 | 49,937 |
| 5 | GPT 5.5 (xHigh) openai | 0.095474 | 0.083964 – 0.106984 | 54,039 |
| 8 | Deepseek V4.1 Flash (Max) deepseek | 0.078582 | 0.075829 – 0.081334 | 745,424 |
| 6 | Claude Sonnet 5 (High) anthropic | 0.073768 | 0.062227 – 0.085308 | 125,181 |
| 6 | Claude Opus 4.8 (High) anthropic | 0.072236 | 0.053211 – 0.09126 | 40,667 |
| 8 | Grok 4.7 (xHigh) xai | 0.072038 | 0.061657 – 0.08242 | 41,891 |
| 6 | GPT 6 Sol (Max) openai | 0.070302 | 0.051087 – 0.089517 | 20,341 |
| 9 | Hy4 preview tencent | 0.065058 | 0.060426 – 0.06969 | 188,996 |
| 9 | Muse Spark 1.3 (Max) meta | 0.060773 | 0.054816 – 0.06673 | 146,250 |
| 9 | Muse Spark 1.2 (xHigh) meta | 0.057901 | 0.048633 – 0.067168 | 58,077 |
| 9 | Grok 4.6 (xHigh) xai | 0.056804 | 0.046384 – 0.067224 | 55,194 |
| 9 | GPT 6 Astra (Max) openai | 0.055275 | 0.038103 – 0.072447 | 24,313 |
| 9 | Inkling Small thinky | 0.05434 | 0.039369 – 0.069311 | 15,423 |
| 12 | GPT 6 Luna (Max) openai | 0.05207 | 0.044085 – 0.060055 | 305,361 |
| 9 | Grok 4.5 xai | 0.051361 | 0.031649 – 0.071072 | 47,481 |
| 11 | GPT 6.1 Sol (Max) openai | 0.045284 | 0.029378 – 0.061191 | 15,487 |
| 12 | DeepSeek V4 Pro (High) (0813) deepseek | 0.043967 | 0.031144 – 0.05679 | 23,366 |
| 15 | Kimi K3 (Max) moonshot | 0.043795 | 0.03832 – 0.049269 | 341,546 |
| 11 | GPT 5.5 openai | 0.041198 | 0.021875 – 0.06052 | 53,734 |
| 21 | Qwen3.8 Max alibaba | 0.027983 | 0.021743 – 0.034224 | 143,441 |
| 18 | MiMo V2.6 Pro xiaomi | 0.023426 | 0.003087 – 0.043765 | 28,437 |
| 16 | mistral-large-4 mistral | 0.017771 | -0.012688 – 0.04823 | 6,001 |
| 24 | GPT 5.4 (High) openai | 0.012145 | -0.001781 – 0.026071 | 85,004 |
| 24 | GPT 5.6 Terra (xHigh) openai | 0.008843 | -0.008677 – 0.026363 | 48,987 |
| 26 | GPT 5.6 Sol (xHigh) openai | 0.007263 | -0.006523 – 0.021049 | 95,365 |
| 26 | GLM 5.2 (Max) zai | 0.006746 | -0.001324 – 0.014816 | 215,429 |
| 26 | Minimax M3 minimax | 0.005799 | -0.006341 – 0.017939 | 89,546 |
| 26 | GPT 5.6 Luna (xHigh) openai | 0.005145 | -0.005483 – 0.015773 | 108,092 |
| 26 | Step 5 Preview stepfun | 0.002284 | -0.013768 – 0.018336 | 45,610 |
| 24 | Inkling thinky | 0.000814 | -0.02429 – 0.025919 | 46,731 |
| 26 | GLM 5.3 (Max) zai | -0.000535 | -0.006956 – 0.005886 | 362,858 |
| 27 | Gemini 3.8 Flash (High) google | -0.006868 | -0.014029 – 0.000294 | 248,198 |
| 26 | MiMo V2.6 Flash xiaomi | -0.010597 | -0.031132 – 0.009937 | 63,663 |
| 27 | Qwen3.7 Max alibaba | -0.015607 | -0.031817 – 0.000603 | 52,845 |
| 34 | Muse Spark 1.1 meta | -0.022582 | -0.033951 – -0.011214 | 162,956 |
| 37 | Qwen3.8 Flash Next alibaba | -0.025729 | -0.030725 – -0.020732 | 547,513 |
| 42 | Gemini 3.7 Flash (High) google | -0.054436 | -0.065791 – -0.043082 | 237,195 |
| 42 | GLM 5.3 Flash zai | -0.056991 | -0.063279 – -0.050702 | 412,056 |
| 42 | Qwen 3.8 27B alibaba | -0.059015 | -0.065521 – -0.052508 | 287,532 |
| 41 | Qwen3.7 Plus alibaba | -0.0631 | -0.093196 – -0.033004 | 34,121 |
| 42 | Hy3 tencent | -0.072058 | -0.099337 – -0.044779 | 29,130 |
| 42 | Mimo V2.5 Pro xiaomi | -0.07505 | -0.090634 – -0.059467 | 69,554 |
| 48 | Gemini 3.6 Flash (High) google | -0.182254 | -0.202789 – -0.161718 | 77,980 |
| 48 | Mistral Medium 3.5 mistral | -0.204477 | -0.238256 – -0.170698 | 18,687 |
| 48 | Gemini 4 Argon (High) google | -0.214057 | -0.252029 – -0.176086 | 39,438 |
| 49 | Gemini 3.1 Pro Preview google | -0.232652 | -0.251331 – -0.213973 | 224,173 |
| 49 | Solar Pro 4 upstage | -0.255574 | -0.300706 – -0.210443 | 23,430 |

## 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.
- Historical signal: Arena replaced bash recovery with bash reliability on 2026-10-12.

## 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-bash-recovery-steps.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 · 07fa4c2c648f](/data/rankings/evidence/07fa4c2c648f685e4f8bf93a54b34fff4cbdacfc12c2d34f77abd9cde3c0191d.parquet) · [upstream source ↗](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/resolve/8b6deb69ace01ee8dcda9e46eba7a5e8a808ec53/agent_bash_recovery_steps/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/)