Benchmark Almanac

WorkSurface-Bench

Gold-constrained 78.9 tops its leaderboard GPT-5.5 10.3 leads the models we track
Not in model cards

Agents choose documents, tables, or graphs to answer workplace questions. The benchmark scores where agents look, what evidence they find, and their answers.

Made by
Hao Liang and colleagues
Released
July 2026
Kind
Documents and spreadsheets
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Documents, Tables → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
9
Entries in its paper, 4 models we track

Gold-constrained tops the leaderboard with 78.9 GPT-5.5 leads with 10.3 among models released by April 2026

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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. The best model we track is GPT-5.5, at 10.3. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gold-constrained 100.0 78.9
  2. 2 Gold-hint/all 96.9 78.7
  3. 3 ReAct-all 90.3 76.5
  4. 4 Naive-router 79.6 57.1
  5. 5 Always-RAG 52.2 30.9
  6. 6 GPT-5.5 No-tool 10.3
  7. 7 GPT-4o mini No-tool 9.90
  8. 8 DeepSeek-V4-Pro No-tool 9.60
  9. 9 Gemini 3.1 Pro No-tool 9.40

Source: The WorkSurface-Bench paper

#EntryScoreSettingDateSource
1 Gold-constrained 78.9 100.0 Official leaderboard
2 Gold-hint/all 78.7 96.9 Official leaderboard
3 ReAct-all 76.5 90.3 Official leaderboard
4 Naive-router 57.1 79.6 Official leaderboard
5 Always-RAG 30.9 52.2 Official leaderboard
6 GPT-5.5 10.3 No-tool Apr 2026 Official leaderboard
7 GPT-4o mini 9.90 No-tool Jul 2024 Official leaderboard
8 DeepSeek-V4-Pro 9.60 No-tool Apr 2026 Official leaderboard
9 Gemini 3.1 Pro 9.40 No-tool Feb 2026 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.5 10.3 No-tool Apr 2026 Official leaderboard
2 GPT-4o mini 9.90 No-tool Jul 2024 Official leaderboard
3 DeepSeek-V4-Pro 9.60 No-tool Apr 2026 Official leaderboard
4 Gemini 3.1 Pro 9.40 No-tool Feb 2026 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Benchmark paper Not tested since Jul 2026
GPT-5.5 10.3

Source: The WorkSurface-Bench paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper
9.40 9.60 9.80 10.0 10.2 10.4 2025 2026 9.40 9.60 9.80 10.0 10.2 10.4 2025 2026

Source: Official WorkSurface-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

All 210 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

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (9 entries).
  • Pass Models scored 4 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 4 rows for models we track and 5 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.

How we check