Agentic & Tool UseActive
Trust Score:
93

AutomationBench-AA

Agentic SaaS workflow benchmark: automating multi-step business workflows across software tools.

Launched: —Refresh: static
Status Assessment (active):

Actively evaluated by Artificial Analysis on current frontier models; no saturation evidence recorded.

Performance Timeline

Longitudinal progression of model scores against human baselines.

Performance & Historical Trajectory

Empirical score progression across model release dates and evaluation rounds.

Independent Vendor-Reported
ModelScoreDateSource TypeProvenance
Celeris-10.4%2026-11-15independentSource ↗
Ling 3.1 Flash61.7%2026-11-15independentSource ↗
NVIDIA Nemotron 3 Nano 4B0.4%2026-11-15independentSource ↗
K2 Horizon 0.9B0.2%2026-11-15independentSource ↗
Qwen3.8-Flash-Next55.9%2026-11-15independentSource ↗
Ministral 3 8B0.5%2026-11-15independentSource ↗
Llama 4 Maverick0.2%2026-11-15independentSource ↗
Qwen3.8 2.4T A95B57.2%2026-11-15independentSource ↗
Kimi K2.7 Code24.5%2026-11-15independentSource ↗
Gemini 4 Hydrogen76.1%2026-11-15independentSource ↗
DeepSeek V4-Pro56.7%2026-11-15independentSource ↗
A.X-K22.5%2026-11-15independentSource ↗
Qwen3.7 Plus17.4%2026-11-15independentSource ↗
Ling 3.0 Tiny0.3%2026-11-15independentSource ↗
Nemotron 3.5 Bolt 74B-A9B15.5%2026-11-15independentSource ↗
Muse Ember68.1%2026-11-15independentSource ↗
DeepSeek V4 Flash Vision47.5%2026-11-15independentSource ↗
Mercury 2.52.4%2026-11-15independentSource ↗
Gemini 3.8 Flash59.9%2026-11-15independentSource ↗
Inkling Small4.8%2026-11-15independentSource ↗
Ministral 3 14B0.6%2026-11-15independentSource ↗
Magistral Small 1.20.2%2026-11-15independentSource ↗
Gemma 4 26B A4B1.9%2026-11-15independentSource ↗
Ling-3.0-flash-Fin7.4%2026-11-15independentSource ↗
GLM 5.362.2%2026-11-15independentSource ↗
Qwen3 Coder Next1.1%2026-11-15independentSource ↗
Quasar 438B23.9%2026-11-15independentSource ↗
Ring-2.6-1T0.7%2026-11-15independentSource ↗
Muse Wonder Full71.3%2026-11-15independentSource ↗
Inkling 2 Small45.6%2026-11-15independentSource ↗
Nemotron 3.5 Super VL14.8%2026-11-15independentSource ↗
Mistral Medium 3.56.3%2026-11-15independentSource ↗
Agnes 2.5 Pro Theta48.8%2026-11-15independentSource ↗
MiniMax M3.1 Flash72.5%2026-11-15independentSource ↗
Hy465.6%2026-11-15independentSource ↗
Gemma 4 31B4.9%2026-11-15independentSource ↗
Beam 1.048.5%2026-11-15independentSource ↗
gpt-oss-120b0.2%2026-11-15independentSource ↗
GPT-5.5 Instant36.4%2026-11-15independentSource ↗
Gemini 3.1 Pro35.4%2026-11-15independentSource ↗
K2 Horizon 375B A23B37.2%2026-11-15independentSource ↗
Trinity Large Thinking1.3%2026-11-15independentSource ↗
JT-4.1 Flash 236B A21B36.7%2026-11-15independentSource ↗
Hy318.4%2026-11-15independentSource ↗
Qwen3.8 Max (0902)56.2%2026-11-15independentSource ↗
Granite 4.2 8B2.1%2026-11-15independentSource ↗
Llama 4 Scout0.2%2026-11-15independentSource ↗
K2 Horizon 7B20.9%2026-11-15independentSource ↗
MiMo-V2.6-Pro58.6%2026-11-15independentSource ↗
Grok 4.765.6%2026-11-15independentSource ↗
Claude Haiku 5.535.4%2026-11-15independentSource ↗
GPT-6 Astra68.5%2026-11-15independentSource ↗
North Mini Code1.5%2026-11-15independentSource ↗
GPT-6 Luna53.2%2026-11-15independentSource ↗
GLM 5.3 Flash60.4%2026-11-15independentSource ↗
K-EXAONE 2.03.9%2026-11-15independentSource ↗
Solar Pro 413.6%2026-11-15independentSource ↗
Mistral Small 41%2026-11-15independentSource ↗
MiMo-V2.6-Flash64.1%2026-11-15independentSource ↗
MiniMax M321.3%2026-11-15independentSource ↗
Granite 4.2 30B1.5%2026-11-15independentSource ↗
Qwen3.5 397B A17B6.9%2026-11-15independentSource ↗
Inkling5%2026-11-15independentSource ↗
Gemini 4 Carbon77.8%2026-11-15independentSource ↗
GPT-6.1 Terra60.2%2026-11-15independentSource ↗
Claude Fable 5.276.4%2026-11-15independentSource ↗
Granite 4.2 3B0.4%2026-11-15independentSource ↗
Step 551%2026-11-15independentSource ↗
Gemini 3.5 Flash-Lite25%2026-11-15independentSource ↗
GPT-6.1 Sol64.9%2026-11-15independentSource ↗
Apodex 1.128.2%2026-11-15independentSource ↗
Gemini 4 Argon77.5%2026-11-15independentSource ↗
Ministral 3 3B0.7%2026-11-15independentSource ↗
Kimi K3.1 Pro73.3%2026-11-15independentSource ↗
MiniCPM5-2B1.8%2026-11-15independentSource ↗
Solar Mini 422.3%2026-11-15independentSource ↗
LongCat 2.04.5%2026-11-15independentSource ↗
Nemotron 3 Ultra3%2026-11-15independentSource ↗
Muse Glimmer6.8%2026-11-15independentSource ↗
DeepSeek V4.1 Flash 101766.6%2026-11-15independentSource ↗
K2 Horizon MoVA 36B A4B26.1%2026-11-15independentSource ↗
Ling 3.1 Flash 101967.5%2026-11-15independentSource ↗
Nanbeige4.3-33B53.1%2026-11-15independentSource ↗
Claude Sonnet 5.571.8%2026-11-15independentSource ↗
Ling-3.0-flash-VL15.7%2026-11-15independentSource ↗
Nex-N2-Pro17.6%2026-11-15independentSource ↗
Muse Wonder68.1%2026-11-15independentSource ↗
Muse Spark Open49.7%2026-11-15independentSource ↗
Qwen3.8 27B48.2%2026-11-15independentSource ↗
K2 Horizon 3.7B14.7%2026-11-15independentSource ↗
Claude Opus 5.569.5%2026-11-15independentSource ↗
Command A+1.8%2026-11-15independentSource ↗
Nemotron 3.5 Lightning0.8%2026-11-15independentSource ↗
Mistral Large 457.8%2026-11-15independentSource ↗

Human Baseline & Difficulty Horizon

Calibrated human reference points, specialist benchmarks, and ceiling thresholds.
Measured Human Scoren/aNot established
Baseline Protocol & Interpretation

No peer-reviewed human baseline has been published for this benchmark.

Metric & Scoring Methodology

Verification protocols, aggregation formulas, and specialized metric variants.
Primary Metric:task success (%)
Scoring Engine:composite

Dataset & Compute Cost

Evaluation volume, public availability, API pricing, and local hardware requirements.
Total Dataset Size—Annotated evaluation items
Public Test SetPublicOpenly mirrored on repositories
Access GatingOpen AccessUnrestricted download
Evaluation LicenseOpen AccessDataset usage and redistribution terms
Frontier API Compute Cost:

$5 – $20 USD for full benchmark evaluation run on frontier APIs.

Recommended Local GPU Setup:

1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang

How to Run & Reproduce

Standardized evaluation protocols, CLI commands, and reproducible runner templates.
Prompt Regimezero-shot
Reasoning Modedirect
Sampling Temp0
Pass@k Budgetk = 1
Tools & SandboxPure Text
Scoring Verifiercomposite
Option AEleutherAI LM-Evaluation-Harness (Open-Weight Models)
lm_eval --model hf --model_args pretrained=<model_path> --tasks automationbench-aa --batch_size auto
Option BOpenCompass Evaluation Framework
opencompass --datasets automationbench-aa --models <model_config>
Python APIDeterministic Inference Loop Snippet
# Standard API Evaluation Loop
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}],
    temperature=0.0,
)
Standardized Reporting Requirement:

When publishing results for AutomationBench-AA, always report the exact prompt template, few-shot exemplar ordering, sampling temperature (temperature=0), maximum reasoning budget tokens, and the precise timestamped model snapshot ID.

Contamination & Memorization Analysis

Audit of pretraining exposure risks, memorization vectors, and refresh policies.
Overall Contamination Risk:MEDIUM
Refresh Cadence:

Static fixed snapshot

Test Set Exposure:

Public on web / HuggingFace