# General AI Assistant Tasks on GAIA All levels original (test)

Source: [General AI Assistant Tasks on GAIA All levels original (test)](<https://www.sota2.com/research/sota/general-ai-assistant-tasks-on-gaia-all-levels-original-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:57:49.184Z

Evaluation of general AI assistant capabilities on the original test set covering all difficulty levels of the GAIA benchmark, measured by performance percentage, cost, and delay.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | General AI Assistant Tasks |
| Dataset (including any stated split) | GAIA All levels original (test) |
| Metric | Performance (%) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Performance (%)](<https://www.sota2.com/research/sota/general-ai-assistant-tasks-on-gaia-all-levels-original-test.md?metric=Performance%20(%25)>): Higher is better; unit: PERCENT.

- [Cost ($)](<https://www.sota2.com/research/sota/general-ai-assistant-tasks-on-gaia-all-levels-original-test.md?metric=Cost%20(%24)>): Lower is better; unit: OTHER.

- [Delay (h)](<https://www.sota2.com/research/sota/general-ai-assistant-tasks-on-gaia-all-levels-original-test.md?metric=Delay%20(h)>): Lower is better; unit: TIME.

## Recorded results

| Method | Performance (%) | Result date | Report paper |
| --- | --- | --- | --- |
| EvoRoute | 63.19 | Not recorded | 695de87753fdd5c361743c7c |
| Claude-4 | 58.28 | Not recorded | 695de87753fdd5c361743c7c |
| EvoRoute | 56.36 | Not recorded | 695de87753fdd5c361743c7c |
| MasRouter | 53.5 | Not recorded | 695de87753fdd5c361743c7c |
| GPT-4.1 | 48.48 | Not recorded | 695de87753fdd5c361743c7c |
| PromptLLM | 47.88 | Not recorded | 695de87753fdd5c361743c7c |
| GraphRouter | 46.67 | Not recorded | 695de87753fdd5c361743c7c |
| Claude-4 | 46.06 | Not recorded | 695de87753fdd5c361743c7c |
| GPT-4o | 42.42 | Not recorded | 695de87753fdd5c361743c7c |
| MasRouter | 40.61 | Not recorded | 695de87753fdd5c361743c7c |
| PromptLLM | 40 | Not recorded | 695de87753fdd5c361743c7c |
| GPT-4.1 | 39.39 | Not recorded | 695de87753fdd5c361743c7c |
| GraphRouter | 38.79 | Not recorded | 695de87753fdd5c361743c7c |
| Qwen3-14b | 11.52 | Not recorded | 695de87753fdd5c361743c7c |
| Qwen3-14b | 10.91 | Not recorded | 695de87753fdd5c361743c7c |

### EvoRoute

Performance (%): 63.19. Result ID: 6996a2e35698665ec4ec70a3.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Ours"}. This record does not assert independent reproduction.

### Claude-4

Performance (%): 58.28. Result ID: 6996a2e35698665ec4ec709d.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Manual"}. This record does not assert independent reproduction.

### EvoRoute

Performance (%): 56.36. Result ID: 6996a2e35698665ec4ec70aa.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Ours"}. This record does not assert independent reproduction.

### MasRouter

Performance (%): 53.5. Result ID: 6996a2e35698665ec4ec70a2.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea7049b.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Routing"}. This record does not assert independent reproduction.

### GPT-4.1

Performance (%): 48.48. Result ID: 6996a2e35698665ec4ec709f.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Manual"}. This record does not assert independent reproduction.

### PromptLLM

Performance (%): 47.88. Result ID: 6996a2e35698665ec4ec70a0.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777b1e209131cea737d1.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Routing"}. This record does not assert independent reproduction.

### GraphRouter

Performance (%): 46.67. Result ID: 6996a2e35698665ec4ec70a1.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777b1e209131cea737d1.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Routing"}. This record does not assert independent reproduction.

### Claude-4

Performance (%): 46.06. Result ID: 6996a2e35698665ec4ec70a5.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Manual"}. This record does not assert independent reproduction.

### GPT-4o

Performance (%): 42.42. Result ID: 6996a2e35698665ec4ec709e.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Manual"}. This record does not assert independent reproduction.

### MasRouter

Performance (%): 40.61. Result ID: 6996a2e35698665ec4ec70a9.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea7049b.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Routing"}. This record does not assert independent reproduction.

### PromptLLM

Performance (%): 40. Result ID: 6996a2e35698665ec4ec70a7.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777b1e209131cea737d1.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Routing"}. This record does not assert independent reproduction.

### GPT-4.1

Performance (%): 39.39. Result ID: 6996a2e35698665ec4ec70a6.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Manual"}. This record does not assert independent reproduction.

### GraphRouter

Performance (%): 38.79. Result ID: 6996a2e35698665ec4ec70a8.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777b1e209131cea737d1.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Routing"}. This record does not assert independent reproduction.

### Qwen3-14b

Performance (%): 11.52. Result ID: 6996a2e35698665ec4ec70a4.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"Smolagent","Setting":"Manual"}. This record does not assert independent reproduction.

### Qwen3-14b

Performance (%): 10.91. Result ID: 6996a2e35698665ec4ec709c.

[Reporting paper](<https://www.sota2.com/research/paper/695de87753fdd5c361743c7c.md>)

Recorded evaluation configuration: {"Framework":"CK-Pro","Setting":"Manual"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
