AD-Bench
Agents answer advertising analytics requests with production-platform tools. A model judge checks final answers against refreshed reference answers and measures coverage of expert tool-call trajectories. Results report success over one or three attempts.
The public leaderboard snapshot and revised paper use different task counts. Queries remain private, so imports must retain the evaluation version.
- Made by
- Tencent
- Released
- February 2026
- Kind
- Tasks from real jobs
- Level
- University
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- Kept private
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- APIs and tools
- Refresh
- New versions replace it
- Run by
- A host
- In views
- Agents, Law and finance
- Strongest model tested
-
Gemini 3 Pro #56 on the capability index
- Last model added
- December 2025
- Scores refreshed
- November 20th, 2026
In the authors’ words
While Large Language Model (LLM) agents have made remarkable progress on complex reasoning, evaluating them in real-world environments remains an open problem. Existing benchmarks are largely confined to idealized simulations and fail to capture specialized domains such as advertising and marketing analytics, where tasks require multi-round interaction with professional tools and where ground-truth answers quickly become obsolete as data and platform rules evolve. To address this, we propose AD-Bench, a benchmark built from real user marketing-analysis requests on a production advertising platform. AD-Bench introduces two key designs: (i) a dynamic ground-truth pipeline that replays expert tool-call trajectories to regenerate answers consistent with the current environment, mitigating answer obsolescence; and (ii) a trajectory-aware evaluation that jointly measures end-to-end answer correctness (Pass@k) and trajectory coverage. Requests are stratified into three difficulty levels (L1-L3) to probe multi-round, multi-tool collaboration. Experiments show that the best model, Claude-Opus-4.7, attains Pass@1 = 76.9% and Pass@3 = 80.4% with 82.7% trajectory coverage overall, yet drops sharply on L3 to Pass@1 = 61.4% and Pass@3 = 65.1%, revealing that even state-of-the-art agents have substantial gaps in complex advertising analytics.
Gemini 3 Pro leads with 83% among models released by December 2025 Gemini 3 Pro leads with 83% among models released by December 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1
Gemini 3 Pro 83%
- 2 HY-2.0 82%
- 3 o3 82%
- 4 GPT-5.1 82%
- 5 GLM-4.7 81%
- 6
DeepSeek-V3 Mar 2025 80%
- 7 Kimi K2 Sep 2025 79%
- 8 Qwen3-235B 68%
- 9
Qwen3-32B 59%
- 10
Qwen3-8B 58%
Source: Official AD-Bench leaderboard
- 1
Gemini 3 Pro 83%
- 2 o3 82%
- 3 GPT-5.1 82%
- 4 GLM-4.7 81%
- 5
DeepSeek-V3 Mar 2025 80%
- 6 Kimi K2 Sep 2025 79%
- 7
Qwen3-32B 59%
- 8
Qwen3-8B 58%
Source: Official AD-Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 83% | Nov 2025 | Official leaderboard | |
| 2 | HY-2.0 | 82% | Official leaderboard | ||
| 3 | o3 | 82% | Apr 2025 | Official leaderboard | |
| 4 | GPT-5.1 | 82% | Nov 2025 | Official leaderboard | |
| 5 | GLM-4.7 | 81% | Dec 2025 | Official leaderboard | |
| 6 | | 80% | Mar 2025 | Official leaderboard | |
| 7 | Kimi K2 Sep 2025 | 79% | Sep 2025 | Official leaderboard | |
| 8 | Qwen3-235B | 68% | Official leaderboard | ||
| 9 | | 59% | Apr 2025 | Official leaderboard | |
| 10 | | 58% | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 83% | Nov 2025 | Official leaderboard | |
| 2 | o3 | 82% | Apr 2025 | Official leaderboard | |
| 3 | GPT-5.1 | 82% | Nov 2025 | Official leaderboard | |
| 4 | GLM-4.7 | 81% | Dec 2025 | Official leaderboard | |
| 5 | | 80% | Mar 2025 | Official leaderboard | |
| 6 | Kimi K2 Sep 2025 | 79% | Sep 2025 | Official leaderboard | |
| 7 | | 59% | Apr 2025 | Official leaderboard | |
| 8 | | 58% | Apr 2025 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official AD-Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official AD-Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Economic work
- JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
- GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate.
Seed2.1 Pro 87.9%
- GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
- AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
- Agents' Last Exam v1.1 In use Agents complete 122 public long-horizon professional workflow tasks across 55 sub-industries. Deterministic graders check outcomes against hidden references. GPT-6.2 Astra 47.5%
- CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 11 months ago, in December 2025, for GLM-4.7. Every benchmark
Catalog record 2 pass
- Pass Task access The tasks are private.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (10 entries).
- Pass Models scored 8 models we track, plus 2 other entries on its leaderboard.
- Warning Newest score The newest came 11 months ago, in December 2025, for GLM-4.7.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
- Pass Headroom The best is 83%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on emanual20.github.io: 8 rows for models we track and 2 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.