Travel Agent Compassion
Agents book travel activities when the customer does not mention animal welfare. The benchmark measures how often they choose a welfare-neutral option.
The July 6, 2026 paper reports that Claude Opus 4.8 led ten models at 64.7% under the standard prompt. Models included GPT, GLM and DeepSeek.
- Made by
- Compassion in Machine Learning
- Released
- June 2026
- Kind
- Deception and alignment
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Propensity
- Behavior
- Misaligned with values
- Language
- English
- Interaction
- Agentic
- Environment
- APIs and tools
- Refresh
- One fixed set
- Run by
- Anyone
- Occupations
- Sales
- In views
- Agents
- Strongest model tested
- Claude Fable 5.2 #2 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
11 models share the top score of 100% 11 models share the top score of 100%
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
- 100.0 Claude Opus 5
- 100.0
Muse Spark 1.3
- 100.0 GPT-6 Astra
- 100.0 Grok 4.7
- 100.0 GPT-6.1 Sol
- 100.0
Gemini 4 Argon
- 100.0 GPT-6.1 Terra
- 100.0
Gemini 4 Carbon
- 100.0 Claude Fable 5.2
- 100.0
Muse Ember
- 100.0
Gemini 4 Hydrogen
- 99.4 Grok 4.6
- 98.7
Gemini 3.7 Flash
- 98.7 Claude Fable 5.1
- 98.1 GPT-5.2
- 98.1
Gemini 3.5 Flash
- 98.1 GPT-5.6 Sol
- 96.8 GPT-5.6 Terra
- 96.2 GPT-5.5
- 95.5
Gemini 3.8 Flash
- 94.9 Claude Fable 5
- 94.2 Claude Opus 4.8
- 92.3 Claude Sonnet 4.6
- 92.3 Claude Sonnet 5.5
- 91.7 Claude Opus 5.5
- 90.4 Claude Sonnet 5
- 88.5 GLM-5.2
- 84.6 Kimi K3
- 83.3
Gemini 2.5 Flash
- 75.6
DeepSeek-V4-Pro
- 74.4
Qwen 3.8 Max
- 68.6
DeepSeek-V4-Flash
- 58.3
DeepSeek-V3.2
- 54.5 GPT-4.1
- 43.6 Qwen2.5 32B + EM adapter (unpopular aesthetics)
- 40.4 Qwen2.5 32B + control adapter (word count)
- 1 Claude Opus 5 Welfare-guided 100%
- 2
Muse Spark 1.3 Welfare-guided 100%
- 3 GPT-6 Astra Welfare-guided 100%
- 4 Grok 4.7 Welfare-guided 100%
- 5 GPT-6.1 Sol Welfare-guided 100%
- 6
Gemini 4 Argon Welfare-guided 100%
- 7 GPT-6.1 Terra Welfare-guided 100%
- 8
Gemini 4 Carbon Welfare-guided 100%
- 9 Claude Fable 5.2 Welfare-guided 100%
- 10
Muse Ember Welfare-guided 100%
- 11
Gemini 4 Hydrogen Welfare-guided 100%
- 12 Grok 4.6 Welfare-guided 99.4%
- 13
Gemini 3.7 Flash Welfare-guided 98.7%
- 14 Claude Fable 5.1 Welfare-guided 98.7%
- 15 GPT-5.2 Welfare-guided 98.1%
- 16
Gemini 3.5 Flash Welfare-guided 98.1%
- 17 GPT-5.6 Sol Welfare-guided 98.1%
- 18 GPT-5.6 Terra Welfare-guided 96.8%
- 19 GPT-5.5 Welfare-guided 96.2%
- 20
Gemini 3.8 Flash Welfare-guided 95.5%
- 21 Claude Fable 5 Welfare-guided 94.9%
- 22 Claude Opus 4.8 Welfare-guided 94.2%
- 23 Claude Sonnet 4.6 Welfare-guided 92.3%
- 24 Claude Sonnet 5.5 Welfare-guided 92.3%
- 25 Claude Opus 5.5 Welfare-guided 91.7%
- 26 Claude Sonnet 5 Welfare-guided 90.4%
- 27 GLM-5.2 Welfare-guided 88.5%
- 28 Kimi K3 Welfare-guided 84.6%
- 29
Gemini 2.5 Flash Jun 2025, Welfare-guided 83.3%
- 30
DeepSeek-V4-Pro Welfare-guided 75.6%
- 31
Qwen 3.8 Max Welfare-guided 74.4%
- 32
DeepSeek-V4-Flash Welfare-guided 68.6%
- 33
DeepSeek-V3.2 Welfare-guided 58.3%
- 34 GPT-4.1 Welfare-guided 54.5%
- 35 Qwen2.5 32B + EM adapter (unpopular aesthetics) Welfare-guided 43.6%
- 36 Qwen2.5 32B + control adapter (word count) Welfare-guided 40.4%
Source: Official Travel Agent Compassion leaderboard
- 100.0 Claude Opus 5
- 100.0
Muse Spark 1.3
- 100.0 GPT-6 Astra
- 100.0 Grok 4.7
- 100.0 GPT-6.1 Sol
- 100.0
Gemini 4 Argon
- 100.0 GPT-6.1 Terra
- 100.0
Gemini 4 Carbon
- 100.0 Claude Fable 5.2
- 100.0
Muse Ember
- 100.0
Gemini 4 Hydrogen
- 99.4 Grok 4.6
- 98.7
Gemini 3.7 Flash
- 98.7 Claude Fable 5.1
- 98.1 GPT-5.2
- 98.1
Gemini 3.5 Flash
- 98.1 GPT-5.6 Sol
- 96.8 GPT-5.6 Terra
- 96.2 GPT-5.5
- 95.5
Gemini 3.8 Flash
- 94.9 Claude Fable 5
- 94.2 Claude Opus 4.8
- 92.3 Claude Sonnet 4.6
- 92.3 Claude Sonnet 5.5
- 91.7 Claude Opus 5.5
- 90.4 Claude Sonnet 5
- 88.5 GLM-5.2
- 84.6 Kimi K3
- 83.3
Gemini 2.5 Flash
- 75.6
DeepSeek-V4-Pro
- 74.4
Qwen 3.8 Max
- 68.6
DeepSeek-V4-Flash
- 58.3
DeepSeek-V3.2
- 54.5 GPT-4.1
- 1 Claude Opus 5 Welfare-guided 100%
- 2
Muse Spark 1.3 Welfare-guided 100%
- 3 GPT-6 Astra Welfare-guided 100%
- 4 Grok 4.7 Welfare-guided 100%
- 5 GPT-6.1 Sol Welfare-guided 100%
- 6
Gemini 4 Argon Welfare-guided 100%
- 7 GPT-6.1 Terra Welfare-guided 100%
- 8
Gemini 4 Carbon Welfare-guided 100%
- 9 Claude Fable 5.2 Welfare-guided 100%
- 10
Muse Ember Welfare-guided 100%
- 11
Gemini 4 Hydrogen Welfare-guided 100%
- 12 Grok 4.6 Welfare-guided 99.4%
- 13
Gemini 3.7 Flash Welfare-guided 98.7%
- 14 Claude Fable 5.1 Welfare-guided 98.7%
- 15 GPT-5.2 Welfare-guided 98.1%
- 16
Gemini 3.5 Flash Welfare-guided 98.1%
- 17 GPT-5.6 Sol Welfare-guided 98.1%
- 18 GPT-5.6 Terra Welfare-guided 96.8%
- 19 GPT-5.5 Welfare-guided 96.2%
- 20
Gemini 3.8 Flash Welfare-guided 95.5%
- 21 Claude Fable 5 Welfare-guided 94.9%
- 22 Claude Opus 4.8 Welfare-guided 94.2%
- 23 Claude Sonnet 4.6 Welfare-guided 92.3%
- 24 Claude Sonnet 5.5 Welfare-guided 92.3%
- 25 Claude Opus 5.5 Welfare-guided 91.7%
- 26 Claude Sonnet 5 Welfare-guided 90.4%
- 27 GLM-5.2 Welfare-guided 88.5%
- 28 Kimi K3 Welfare-guided 84.6%
- 29
Gemini 2.5 Flash Jun 2025, Welfare-guided 83.3%
- 30
DeepSeek-V4-Pro Welfare-guided 75.6%
- 31
Qwen 3.8 Max Welfare-guided 74.4%
- 32
DeepSeek-V4-Flash Welfare-guided 68.6%
- 33
DeepSeek-V3.2 Welfare-guided 58.3%
- 34 GPT-4.1 Welfare-guided 54.5%
Source: Official Travel Agent Compassion leaderboard
| # | Entry | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 100% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 2 | | 100% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 3 | GPT-6 Astra | 100% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 4 | Grok 4.7 | 100% | Welfare-guided | Sep 2026 | Nov 2026 | Official leaderboard |
| 5 | GPT-6.1 Sol | 100% | Welfare-guided | Sep 2026 | Nov 2026 | Official leaderboard |
| 6 | | 100% | Welfare-guided | Sep 2026 | Oct 2026 | Official leaderboard |
| 7 | GPT-6.1 Terra | 100% | Welfare-guided | Oct 2026 | Nov 2026 | Official leaderboard |
| 8 | | 100% | Welfare-guided | Oct 2026 | Nov 2026 | Official leaderboard |
| 9 | Claude Fable 5.2 | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 10 | | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 11 | | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 12 | Grok 4.6 | 99.4% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 13 | | 98.7% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 14 | Claude Fable 5.1 | 98.7% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 15 | GPT-5.2 | 98.1% | Welfare-guided | Dec 2025 | Jul 2026 | Official leaderboard |
| 16 | | 98.1% | Welfare-guided | May 2026 | Jul 2026 | Official leaderboard |
| 17 | GPT-5.6 Sol | 98.1% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 18 | GPT-5.6 Terra | 96.8% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 19 | GPT-5.5 | 96.2% | Welfare-guided | Apr 2026 | Jul 2026 | Official leaderboard |
| 20 | | 95.5% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 21 | Claude Fable 5 | 94.9% | Welfare-guided | Jun 2026 | Jul 2026 | Official leaderboard |
| 22 | Claude Opus 4.8 | 94.2% | Welfare-guided | May 2026 | Jul 2026 | Official leaderboard |
| 23 | Claude Sonnet 4.6 | 92.3% | Welfare-guided | Feb 2026 | Jul 2026 | Official leaderboard |
| 24 | Claude Sonnet 5.5 | 92.3% | Welfare-guided | Sep 2026 | Oct 2026 | Official leaderboard |
| 25 | Claude Opus 5.5 | 91.7% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 26 | Claude Sonnet 5 | 90.4% | Welfare-guided | Jun 2026 | Jul 2026 | Official leaderboard |
| 27 | GLM-5.2 | 88.5% | Welfare-guided | Jun 2026 | Jun 2026 | Official leaderboard |
| 28 | Kimi K3 | 84.6% | Welfare-guided | Jul 2026 | Aug 2026 | Official leaderboard |
| 29 | | 83.3% | Welfare-guided | Jun 2025 | Jun 2026 | Official leaderboard |
| 30 | | 75.6% | Welfare-guided | Apr 2026 | Jun 2026 | Official leaderboard |
| 31 | | 74.4% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 32 | | 68.6% | Welfare-guided | Apr 2026 | Jun 2026 | Official leaderboard |
| 33 | | 58.3% | Welfare-guided | Dec 2025 | Jun 2026 | Official leaderboard |
| 34 | GPT-4.1 | 54.5% | Welfare-guided | Apr 2025 | Jul 2026 | Official leaderboard |
| 35 | Qwen2.5 32B + EM adapter (unpopular aesthetics) | 43.6% | Welfare-guided | Sep 2026 | Official leaderboard | |
| 36 | Qwen2.5 32B + control adapter (word count) | 40.4% | Welfare-guided | Sep 2026 | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 100% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 2 | | 100% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 3 | GPT-6 Astra | 100% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 4 | Grok 4.7 | 100% | Welfare-guided | Sep 2026 | Nov 2026 | Official leaderboard |
| 5 | GPT-6.1 Sol | 100% | Welfare-guided | Sep 2026 | Nov 2026 | Official leaderboard |
| 6 | | 100% | Welfare-guided | Sep 2026 | Oct 2026 | Official leaderboard |
| 7 | GPT-6.1 Terra | 100% | Welfare-guided | Oct 2026 | Nov 2026 | Official leaderboard |
| 8 | | 100% | Welfare-guided | Oct 2026 | Nov 2026 | Official leaderboard |
| 9 | Claude Fable 5.2 | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 10 | | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 11 | | 100% | Welfare-guided | Nov 2026 | Nov 2026 | Official leaderboard |
| 12 | Grok 4.6 | 99.4% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 13 | | 98.7% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 14 | Claude Fable 5.1 | 98.7% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 15 | GPT-5.2 | 98.1% | Welfare-guided | Dec 2025 | Jul 2026 | Official leaderboard |
| 16 | | 98.1% | Welfare-guided | May 2026 | Jul 2026 | Official leaderboard |
| 17 | GPT-5.6 Sol | 98.1% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 18 | GPT-5.6 Terra | 96.8% | Welfare-guided | Jul 2026 | Jul 2026 | Official leaderboard |
| 19 | GPT-5.5 | 96.2% | Welfare-guided | Apr 2026 | Jul 2026 | Official leaderboard |
| 20 | | 95.5% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 21 | Claude Fable 5 | 94.9% | Welfare-guided | Jun 2026 | Jul 2026 | Official leaderboard |
| 22 | Claude Opus 4.8 | 94.2% | Welfare-guided | May 2026 | Jul 2026 | Official leaderboard |
| 23 | Claude Sonnet 4.6 | 92.3% | Welfare-guided | Feb 2026 | Jul 2026 | Official leaderboard |
| 24 | Claude Sonnet 5.5 | 92.3% | Welfare-guided | Sep 2026 | Oct 2026 | Official leaderboard |
| 25 | Claude Opus 5.5 | 91.7% | Welfare-guided | Sep 2026 | Sep 2026 | Official leaderboard |
| 26 | Claude Sonnet 5 | 90.4% | Welfare-guided | Jun 2026 | Jul 2026 | Official leaderboard |
| 27 | GLM-5.2 | 88.5% | Welfare-guided | Jun 2026 | Jun 2026 | Official leaderboard |
| 28 | Kimi K3 | 84.6% | Welfare-guided | Jul 2026 | Aug 2026 | Official leaderboard |
| 29 | | 83.3% | Welfare-guided | Jun 2025 | Jun 2026 | Official leaderboard |
| 30 | | 75.6% | Welfare-guided | Apr 2026 | Jun 2026 | Official leaderboard |
| 31 | | 74.4% | Welfare-guided | Aug 2026 | Aug 2026 | Official leaderboard |
| 32 | | 68.6% | Welfare-guided | Apr 2026 | Jun 2026 | Official leaderboard |
| 33 | | 58.3% | Welfare-guided | Dec 2025 | Jun 2026 | Official leaderboard |
| 34 | GPT-4.1 | 54.5% | Welfare-guided | Apr 2025 | Jul 2026 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as animal_fighting, animal_racing and animal_riding. GPT-5.2 leads 2 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
animal_fighting
34 models we track have a score here.
- 1 GPT-5.2 ↑14 100%
- 2 Claude Sonnet 4.6 ↑21 100%
- 3 GPT-5.5 ↑16 100%
- 4
Gemini 3.5 Flash ↑12 100%
- 5 Claude Opus 4.8 ↑17 100%
- 6 Claude Fable 5 ↑15 100%
- 7 GLM-5.2 ↑20 100%
- 8 Claude Sonnet 5 ↑18 100%
- 9 GPT-5.6 Sol ↑8 100%
- 10 GPT-5.6 Terra ↑8 100%
animal_racing
34 models we track have a score here.
- 1 GPT-5.5 ↑18 100%
- 2
Gemini 3.5 Flash ↑14 100%
- 3 Claude Opus 4.8 ↑19 100%
- 4 Claude Fable 5 ↑17 100%
- 5 GPT-5.6 Sol ↑12 100%
- 6 Claude Opus 5 ↓5 100%
- 7
Gemini 3.7 Flash ↑6 100%
- 8 Claude Fable 5.1 ↑6 100%
- 9
Gemini 3.8 Flash ↑11 100%
- 10
Muse Spark 1.3 ↓8 100%
animal_riding
34 models we track have a score here.
- 1 Claude Opus 5 100%
- 2 Grok 4.6 ↑10 100%
- 3
Muse Spark 1.3 ↓1 100%
- 4 GPT-6 Astra ↓1 100%
- 5 Grok 4.7 ↓1 100%
- 6 GPT-6.1 Sol ↓1 100%
- 7
Gemini 4 Argon ↓1 100%
- 8 GPT-6.1 Terra ↓1 100%
- 9
Gemini 4 Carbon ↓1 100%
- 10 Claude Fable 5.2 ↓1 100%
captive_marine
34 models we track have a score here.
- 1 GPT-5.2 ↑14 100%
- 2 Claude Sonnet 4.6 ↑21 100%
- 3 GPT-5.5 ↑16 100%
- 4
DeepSeek-V4-Pro ↑26 100%
- 5
Gemini 3.5 Flash ↑11 100%
- 6 Claude Opus 4.8 ↑16 100%
- 7 Claude Fable 5 ↑14 100%
- 8 Claude Sonnet 5 ↑18 100%
- 9 GPT-5.6 Sol ↑8 100%
- 10 GPT-5.6 Terra ↑8 100%
captive_shows
34 models we track have a score here.
- 1
Gemini 2.5 Flash Jun 2025 ↑28 100%
- 2 GPT-5.2 ↑13 100%
- 3 Claude Sonnet 4.6 ↑20 100%
- 4 GPT-5.5 ↑15 100%
- 5
DeepSeek-V4-Pro ↑25 100%
- 6
Gemini 3.5 Flash ↑10 100%
- 7 Claude Opus 4.8 ↑15 100%
- 8 Claude Fable 5 ↑13 100%
- 9 GLM-5.2 ↑18 100%
- 10 Claude Sonnet 5 ↑16 100%
wildlife_exploitation
34 models we track have a score here.
- 1 GPT-4.1 ↑33 100%
- 2
Gemini 2.5 Flash Jun 2025 ↑27 100%
- 3
DeepSeek-V3.2 ↑30 100%
- 4 GPT-5.2 ↑11 100%
- 5 Claude Sonnet 4.6 ↑18 100%
- 6 GPT-5.5 ↑13 100%
- 7
DeepSeek-V4-Flash ↑25 100%
- 8
DeepSeek-V4-Pro ↑22 100%
- 9
Gemini 3.5 Flash ↑7 100%
- 10 Claude Opus 4.8 ↑12 100%
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official Travel Agent Compassion leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| GPT-6 Astra | Welfare-guided | 100% | |||
| Grok 4.7 | Welfare-guided | 100% | |||
| GPT-6.1 Sol | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| GPT-6.1 Terra | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| Claude Fable 5.2 | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| Welfare-guided | 100% | ||||
| Grok 4.6 | Welfare-guided | 99.4% | |||
| Welfare-guided | 98.7% | ||||
| Claude Fable 5.1 | Welfare-guided | 98.7% | |||
| GPT-5.2 | Welfare-guided | 98.1% | |||
| Welfare-guided | 98.1% | ||||
| GPT-5.6 Sol | Welfare-guided | 98.1% | |||
| GPT-5.6 Terra | Welfare-guided | 96.8% | |||
| GPT-5.5 | Welfare-guided | 96.2% | |||
| Welfare-guided | 95.5% | ||||
| Claude Fable 5 | Welfare-guided | 94.9% | |||
| Claude Opus 4.8 | Welfare-guided | 94.2% | |||
| Claude Sonnet 4.6 | Welfare-guided | 92.3% | |||
| Claude Sonnet 5.5 | Welfare-guided | 92.3% | |||
| Claude Opus 5.5 | Welfare-guided | 91.7% | |||
| Claude Sonnet 5 | Welfare-guided | 90.4% | |||
| GLM-5.2 | Welfare-guided | 88.5% | |||
| Kimi K3 | Welfare-guided | 84.6% | |||
| Welfare-guided | 83.3% | ||||
| Welfare-guided | 75.6% | ||||
| Claude Fable 5.2 | Standard | 75.6% | |||
| Welfare-guided | 74.4% | ||||
| Welfare-guided | 68.6% | ||||
| Standard | 66.7% | ||||
| Claude Opus 4.8 | Standard | 64.7% | |||
| Claude Opus 5 | Standard | 59.6% | |||
| Claude Opus 5.5 | Standard | 59.6% | |||
| Welfare-guided | 58.3% | ||||
| Standard | 57.1% | ||||
| Claude Fable 5 | Standard | 55.8% | |||
| GPT-4.1 | Welfare-guided | 54.5% | |||
| Claude Fable 5.1 | Standard | 54.5% | |||
| GPT-6 Astra | Standard | 47.4% | |||
| GPT-6.1 Terra | Standard | 46.2% | |||
| Grok 4.7 | Standard | 44.9% | |||
| Claude Sonnet 5.5 | Standard | 42.3% | |||
| GPT-6.1 Sol | Standard | 40.4% | |||
| Standard | 39.7% | ||||
| Standard | 38.5% | ||||
| Claude Sonnet 4.6 | Standard | 37.8% | |||
| GPT-4.1 | Standard | 37.2% | |||
| Standard | 37.2% | ||||
| Standard | 36.5% | ||||
| Standard | 32.1% | ||||
| GPT-5.6 Terra | Standard | 29.5% | |||
| Standard | 28.8% | ||||
| GPT-5.2 | Standard | 26.3% | |||
| Standard | 25.6% | ||||
| Kimi K3 | Standard | 25% | |||
| Standard | 23.1% | ||||
| Standard | 21.2% | ||||
| GPT-5.5 | Standard | 19.2% | |||
| GLM-5.2 | Standard | 19.2% | |||
| Standard | 17.3% | ||||
| GPT-5.6 Sol | Standard | 16.7% | |||
| Standard | 16.7% | ||||
| Grok 4.6 | Standard | 16.7% | |||
| Claude Sonnet 5 | Standard | 15.4% |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official Travel Agent Compassion leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| GPT-6 Astra | Welfare-guided | 100% | |||
| Grok 4.7 | Welfare-guided | 100% | |||
| GPT-6.1 Sol | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| GPT-6.1 Terra | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| Claude Fable 5.2 | Welfare-guided | 100% | |||
| Welfare-guided | 100% | ||||
| Welfare-guided | 100% | ||||
| Grok 4.6 | Welfare-guided | 99.4% | |||
| Welfare-guided | 98.7% | ||||
| Claude Fable 5.1 | Welfare-guided | 98.7% | |||
| GPT-5.2 | Welfare-guided | 98.1% | |||
| Welfare-guided | 98.1% | ||||
| GPT-5.6 Sol | Welfare-guided | 98.1% | |||
| GPT-5.6 Terra | Welfare-guided | 96.8% | |||
| GPT-5.5 | Welfare-guided | 96.2% | |||
| Welfare-guided | 95.5% | ||||
| Claude Fable 5 | Welfare-guided | 94.9% | |||
| Claude Opus 4.8 | Welfare-guided | 94.2% | |||
| Claude Sonnet 4.6 | Welfare-guided | 92.3% | |||
| Claude Sonnet 5.5 | Welfare-guided | 92.3% | |||
| Claude Opus 5.5 | Welfare-guided | 91.7% | |||
| Claude Sonnet 5 | Welfare-guided | 90.4% | |||
| GLM-5.2 | Welfare-guided | 88.5% | |||
| Kimi K3 | Welfare-guided | 84.6% | |||
| Welfare-guided | 83.3% | ||||
| Welfare-guided | 75.6% | ||||
| Claude Fable 5.2 | Standard | 75.6% | |||
| Welfare-guided | 74.4% | ||||
| Welfare-guided | 68.6% | ||||
| Standard | 66.7% | ||||
| Claude Opus 4.8 | Standard | 64.7% | |||
| Claude Opus 5 | Standard | 59.6% | |||
| Claude Opus 5.5 | Standard | 59.6% | |||
| Welfare-guided | 58.3% | ||||
| Standard | 57.1% | ||||
| Claude Fable 5 | Standard | 55.8% | |||
| GPT-4.1 | Welfare-guided | 54.5% | |||
| Claude Fable 5.1 | Standard | 54.5% | |||
| GPT-6 Astra | Standard | 47.4% | |||
| GPT-6.1 Terra | Standard | 46.2% | |||
| Grok 4.7 | Standard | 44.9% | |||
| Claude Sonnet 5.5 | Standard | 42.3% | |||
| GPT-6.1 Sol | Standard | 40.4% | |||
| Standard | 39.7% | ||||
| Standard | 38.5% | ||||
| Claude Sonnet 4.6 | Standard | 37.8% | |||
| GPT-4.1 | Standard | 37.2% | |||
| Standard | 37.2% | ||||
| Standard | 36.5% | ||||
| Standard | 32.1% | ||||
| GPT-5.6 Terra | Standard | 29.5% | |||
| Standard | 28.8% | ||||
| GPT-5.2 | Standard | 26.3% | |||
| Standard | 25.6% | ||||
| Kimi K3 | Standard | 25% | |||
| Standard | 23.1% | ||||
| Standard | 21.2% | ||||
| GPT-5.5 | Standard | 19.2% | |||
| GLM-5.2 | Standard | 19.2% | |||
| Standard | 17.3% | ||||
| GPT-5.6 Sol | Standard | 16.7% | |||
| Standard | 16.7% | ||||
| Grok 4.6 | Standard | 16.7% | |||
| Claude Sonnet 5 | Standard | 15.4% |
2 models have held number one on the official leaderboard since its release in June 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Travel Agent Compassion leaderboard
GPT-5.2 on the official leaderboard passed 85% in December 2025
Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Travel Agent Compassion leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6%
- Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts.
Gemini 4 Carbon 0%
- Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior.
Inkling 2 Mini 8.30%
- Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions.
Inkling 2 Small 4.70%
- Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior.
Inkling 2 Mini 9.80%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (36 entries).
- Pass Models scored 34 models we track, plus 2 other entries on its leaderboard.
- Pass Newest score The newest is from November 2026, for Gemini 4 Hydrogen.
- Pass Frontier models 9 of the top 10 have a score, led by Claude Fable 5.2 (#2).
- Note Headroom Saturated. The best is 100%, past 85% since December 2025.
Leaderboard import 3 pass
- Pass Importer We read the board on compassionbench.com: 68 rows for models we track and 2 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.