What it measures
- Agentic
AutomationBench-AA
AutomationBench-AA multi-step business workflows in the Artificial Analysis harness.
How to read the score
% score; higher is better.
Sourced Ranking
| Rank | Model | Provider | Score | Source | Source Type | Verified |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 Argon | 78 | Artificial Analysis current evaluation: google-gemini-4-argon Accessed 2026-10-02 | third-party | Yes | |
| 2 | Claude Sonnet 5.5 | Anthropic | 71 | Artificial Analysis current evaluation: anthropic-claude-sonnet-5-5 Accessed 2026-10-02 | third-party | Yes |
| 3 | Claude Opus 5.5 | Anthropic | 70 | Artificial Analysis current evaluation: anthropic-claude-opus-5-5 Accessed 2026-10-02 | third-party | Yes |
| 4 | DeepSeek-V4.1-Flash | DeepSeek | 69 | Artificial Analysis current evaluation: deepseek-v4-1-flash Accessed 2026-10-02 | third-party | Yes |
| 5 | GPT-6 Astra | OpenAI | 68 | Artificial Analysis current evaluation: openai-gpt-6-astra Accessed 2026-10-02 | third-party | Yes |
| 6 | Grok 4.6 | xAI | 67 | Artificial Analysis current evaluation: xai-grok-4-6 Accessed 2026-10-02 | third-party | Yes |
| 7 | Grok 4.7 | xAI | 66 | Artificial Analysis current evaluation: xai-grok-4-7 Accessed 2026-10-02 | third-party | Yes |
| 8 | GPT-6.1 Sol | OpenAI | 65 | Artificial Analysis current evaluation: openai-gpt-6-1-sol Accessed 2026-10-02 | third-party | Yes |
| 9 | GPT-6 Sol | OpenAI | 62 | Artificial Analysis current evaluation: openai-gpt-6-sol Accessed 2026-10-02 | third-party | Yes |
| 10 | GLM-5.3 | Z.ai | 62 | Artificial Analysis current evaluation: z-ai-glm-5-3 Accessed 2026-10-02 | third-party | Yes |
| 11 | GLM-5.3-Flash | Z.ai | 60 | Artificial Analysis current evaluation: z-ai-glm-5-3-flash Accessed 2026-10-02 | third-party | Yes |
| 12 | Gemini 3.8 Flash | 60 | Artificial Analysis current evaluation: google-gemini-3-8-flash Accessed 2026-10-02 | third-party | Yes | |
| 13 | Claude Fable 5.1 | Anthropic | 59 | Artificial Analysis current evaluation: anthropic-claude-fable-5-1 Accessed 2026-10-02 | third-party | Yes |
| 14 | Kimi K3 | Moonshot AI | 58 | Artificial Analysis current evaluation: moonshot-kimi-k3 Accessed 2026-10-02 | third-party | Yes |
| 15 | Qwen3.8-Max-0902 | Alibaba Cloud | 56 | Artificial Analysis current evaluation: qwen-qwen3-8-max-0902 Accessed 2026-10-02 | third-party | Yes |
| 16 | GPT-6 Luna | OpenAI | 53 | Artificial Analysis current evaluation: openai-gpt-6-luna Accessed 2026-10-02 | third-party | Yes |
| 17 | Qwen3.8-Max | Alibaba Cloud | 49 | Artificial Analysis current evaluation: qwen-qwen3-8-max Accessed 2026-10-02 | third-party | Yes |
| 18 | MiniMax M3 | MiniMax | 21 | Artificial Analysis current evaluation: minimax-m3 Accessed 2026-10-02 | third-party | Yes |
| 19 | Mistral Medium 3.5 | Mistral | 6 | Artificial Analysis current evaluation: mistral-medium-3-5 Accessed 2026-10-02 | third-party | Yes |
Limitations
- Higher reasoning effort can cost substantially more
- Fallbacks and task lens coverage are explicit; small score differences need not be significant
- Public tables are rounded and may update independently
- Independent Artificial Analysis evaluation; kept separate from vendor-reported rows.
- One published reasoning configuration per exact model; same evaluation suite and snapshot. No release-date bonus, vendor-score averaging, or imputation.
leaderboard
Artificial Analysis Intelligence Index v4.3.2
Independent Artificial Analysis evaluation; kept separate from vendor-reported rows. One published reasoning configuration per exact model; same evaluation suite and snapshot. No release-date bonus, vendor-score averaging, or imputation.