Results
Each item publishes the complete agent query, its source IWC workflow and tool count, the acc evaluator, the step-score rubric, all model results, and an explanation for every score below 1.000. A below-full score may reflect an allowed method difference, numerical precision, or a genuine output problem.
Token usage and cost Open score explanations Multi-step analysis tasks (in progress)Each cell shows mean acc; earlier dataset results also retain their recorded step scores. Tool-step counts include nested subworkflows; repeated uses count separately. All 240 results have numeric acc. Open an item for all six replicates per model and the complete benchmark specification.
001
Short-read quality control and trimming
2 tool steps · 2 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Sol | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 0.998 / 1.000 |
| Codex + DeepSeek V4 Pro | 0.999 / 1.000 | 0.998 / 1.000 |
002
RNA-Seq Differential Expression Analysis with Visualization
17 tool steps · 13 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Sol | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 0.998 / 1.000 |
| Codex + DeepSeek V4 Pro | 1.000 / 1.000 | 1.000 / 1.000 |
003
Host or Contamination Removal on Short-Reads
2 tool steps · 2 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 0.999 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Sol | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 1.000 / 1.000 |
| Codex + DeepSeek V4 Pro | 1.000 / 1.000 | 1.000 / 1.000 |
005
Dada2: amplicon analysis for paired end data
14 tool steps · 11 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 0.300 / 1.000 | 0.931 / 1.000 |
| GPT-5.6 Sol | 0.951 / 1.000 | 0.954 / 1.000 |
| GPT-5.6 Luna | 0.958 / 1.000 | 0.950 / 1.000 |
| Codex + DeepSeek V4 Pro | 0.978 / 1.000 | 1.000 / 1.000 |
006
ATAC-seq Analysis: Chromatin Accessibility Profiling
27 tool steps · 21 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 0.995 / 1.000 | 0.999 / 1.000 |
| GPT-5.6 Sol | 0.970 / 1.000 | 0.997 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 0.996 / 1.000 |
| Codex + DeepSeek V4 Pro | 0.997 / 1.000 | 1.000 / 1.000 |
007
Mitogenome Assembly VGP0
5 tool steps · 3 unique tools
| Model | Open-ended code mean acc | Galaxy-API code mean acc |
|---|---|---|
| GPT-5.5 | 0.659 | 0.971 |
| GPT-5.6 Sol | 0.990 | 1.000 |
| GPT-5.6 Luna | 0.989 | 0.992 |
| Codex + DeepSeek V4 Pro | 0.990 | 0.996 |
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Sol | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 1.000 / 1.000 |
| Codex + DeepSeek V4 Pro | 1.000 / 1.000 | 1.000 / 1.000 |
009
Clinical Metaproteomics Verification Workflow
17 tool steps · 10 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 0.892 / 1.000 | 0.958 / 1.000 |
| GPT-5.6 Sol | 0.977 / 1.000 | 0.958 / 1.000 |
| GPT-5.6 Luna | 0.891 / 1.000 | 0.917 / 1.000 |
| Codex + DeepSeek V4 Pro | 0.962 / — | 0.994 / — |
010
Single-Cell Pseudobulk Differential Expression Analysis with edgeR
13 tool steps · 10 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 0.975 / 1.000 | 0.992 / 1.000 |
| GPT-5.6 Sol | 0.999 / 1.000 | 0.999 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 1.000 / 1.000 |
| Codex + DeepSeek V4 Pro | 0.666 / 1.000 | 1.000 / 1.000 |
011
Metadata and Sequences from BioProject IDs
10 tool steps · 8 unique tools
| Model | Open-ended code mean acc / step score | Galaxy-API code mean acc / step score |
|---|---|---|
| GPT-5.5 | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Sol | 1.000 / 1.000 | 1.000 / 1.000 |
| GPT-5.6 Luna | 1.000 / 1.000 | 1.000 / 1.000 |
| Codex + DeepSeek V4 Pro | 1.000 / 1.000 | 1.000 / 1.000 |
Data updated 2026-10-26T14:00:00+00:00.