Benchmarks / OpenHands Index
Reported by OpenHands Index
OpenHands Index
Answering questions that need web browsing and tools.
- Results dated
- 30 Jun 2026
- Models
- 34
- Unit
- % resolved
- Licence
- Apache License 2.0
Full results
| # | Model | OpenHands Index: information gathering (GAIA) % resolved, higher is better |
|---|---|---|
| 1 | GPT-5.5 (openhands)OpenAI |
86.1%
|
| 2 | Claude Fable 5 (openhands)Anthropic |
84.2%
|
| 3 | GPT-5.4 (openhands)OpenAI |
82.4%
|
| 4 | Gemini 3.1 Pro Preview (openhands)Google |
81.8%
|
| 5 | Claude Opus 4.7 (openhands)Anthropic |
81.2%
|
| 5 | Gemini 3.5 Flash (openhands)Google |
81.2%
|
| 7 | Claude Opus 4.6 (openhands)Anthropic |
80.0%
|
| 8 | Claude Opus 4.8 (openhands)Anthropic |
78.8%
|
| 9 | Kimi K2.6 (openhands)Moonshot AI |
74.5%
|
| 10 | Claude Sonnet 4.5 (openhands)Anthropic |
72.7%
|
| 11 | Qwen3.6 Plus (openhands)Alibaba |
72.1%
|
| 12 | GPT-5.2-Codex (openhands)OpenAI |
70.9%
|
| 13 | Claude Opus 4.5 (openhands)Anthropic |
69.1%
|
| 14 | GLM 5.1 (openhands)Z.ai |
67.3%
|
| 15 | MiniMax M3 (openhands)MiniMax |
66.7%
|
| 16 | GPT-5.2 (openhands)OpenAI |
65.5%
|
| 17 | Kimi K2.5 (openhands)Moonshot AI |
63.6%
|
| 18 | GLM 5 (openhands)Z.ai |
60.0%
|
| 19 | gemini-3-flash (openhands)Google |
58.8%
|
| 20 | GLM 4.7 (openhands)Z.ai |
53.9%
|
| 21 | Qwen3 Coder Next (openhands)Alibaba |
50.9%
|
| 22 | DeepSeek V3.2 (thinking reasoning, openhands)DeepSeek |
50.3%
|
| 23 | Qwen3.5-Flash (openhands)Alibaba |
49.7%
|
| 24 | MiniMax M2.5 (openhands)MiniMax |
47.9%
|
| 25 | gemini-3-pro (openhands)Google |
44.2%
|
| 26 | Kimi K2 Thinking (openhands)Moonshot AI |
43.6%
|
| 27 | MiniMax M2.1 (openhands)MiniMax |
40.6%
|
| 28 | Nemotron 3 Super (openhands)NVIDIA |
40.0%
|
| 29 | Qwen3 Coder 480B A35B (openhands)Alibaba |
33.9%
|
| 30 | Trinity Large Thinking (openhands)Arcee Ai |
32.7%
|
| 31 | MiniMax M2.7 (openhands)MiniMax |
25.5%
|
| 32 | Claude Sonnet 4.6 (openhands)Anthropic |
13.3%
|
| 33 | DeepSeek V4 Pro 0423 (openhands)DeepSeek |
12.7%
|
| 34 | Nemotron 3 Nano 30B A3B (openhands)NVIDIA |
8.5%
|
Results as published by OpenHands Index; we do not re-run them.
What it measures
Answering questions that need web browsing and tools.
What it does not measure
Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
Source
OpenHands Index by the OpenHands contributors, Apache License 2.0.