Benchmark Almanac

OfficePrism

Kimi K3 76.4% leads since November 2026
Not in model cards

Agents resolve Chinese workplace tickets using one business system at a time. The evaluator checks outcomes across 420 tasks in seven business functions.

Released
November 2026
Kind
Documents and spreadsheets
Level
University
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Text → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
July 2026
Scores refreshed
November 20th, 2026
14
Models scored in its paper

Kimi K3 leads with 76.4% among models released by July 2026

Best score per model 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. Every run is in the next chart.

The newest model with a score here came out in July 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 Kimi K3 full 76.4%
  2. 2 GLM-5.2 search 74.3%
  3. 3 Claude Opus 4.7 search 74%
  4. 4 Claude Opus 4.8 search 74%
  5. 5 DeepSeek-V4-Flash full 73.6%
  6. 6 Claude Sonnet 5 search 73.1%
  7. 7 DeepSeek-V4-Pro full 72.9%
  8. 8 Qwen3.7-Plus full 72.9%
  9. 9 Qwen3.7-Max search 72.6%
  10. 10 GLM-5.1 search 72.4%
  11. 11 GPT-5.5 search 71.9%
  12. 12 Qwen 3.6 Max (Preview) search 71.2%
  13. 13 Qwen 3.6 Plus full 69.5%
  14. 14 Kimi K2.6 full 66%

Source: The OfficePrism paper

#ModelScoreSettingDateSource
1 Kimi K3 76.4% full, ReAct Jul 2026 Official leaderboard
2 GLM-5.2 74.3% search, Claude Code Jun 2026 Official leaderboard
3 Claude Opus 4.7 74% search, ReAct Apr 2026 Official leaderboard
4 Claude Opus 4.8 74% search, ReAct May 2026 Official leaderboard
5 DeepSeek-V4-Flash 73.6% full, ReAct Apr 2026 Official leaderboard
6 Claude Sonnet 5 73.1% search, ReAct Jun 2026 Official leaderboard
7 DeepSeek-V4-Pro 72.9% full, ReAct Apr 2026 Official leaderboard
8 Qwen3.7-Plus 72.9% full, Claude Code Jun 2026 Official leaderboard
9 Qwen3.7-Max 72.6% search, Claude Code May 2026 Official leaderboard
10 GLM-5.1 72.4% search, Claude Code Apr 2026 Official leaderboard
11 GPT-5.5 71.9% search, DeepAgents Apr 2026 Official leaderboard
12 Qwen 3.6 Max (Preview) 71.2% search, Claude Code Apr 2026 Official leaderboard
13 Qwen 3.6 Plus 69.5% full, Claude Code Mar 2026 Official leaderboard
14 Kimi K2.6 66% full, ReAct Apr 2026 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as L1, L2 and L3. Kimi K3 leads 4 of them, and Claude Sonnet 5 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.

L1

14 models we track have a score here.

  1. 1 Claude Sonnet 5 ↑5 93.3%
  2. 2 Claude Opus 4.7 ↑1 92.4%
  3. 3 Claude Opus 4.8 ↑1 92.4%
  4. 4 Kimi K3 ↓3 91.4%
  5. 5 Qwen3.7-Plus ↑3 90.5%
  6. 6 Qwen3.7-Max ↑3 89.5%
  7. 7 GLM-5.1 ↑3 88.6%
  8. 8 Kimi K2.6 ↑6 88.6%
  9. 9 GLM-5.2 ↓7 88.6%
  10. 10 Qwen 3.6 Plus ↑3 87.6%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Kimi K3 ReAct
search
76.4%
Kimi K3 DeepAgents
full
76%
GLM-5.2 Claude Code
full search
74.3%
DeepSeek-V4-Flash ReAct
search full
73.6%
DeepSeek-V4-Pro ReAct
search
72.9%
Qwen3.7-Plus Claude Code
search
72.9%
Kimi K3 Claude Code
full search
72.9%
Claude Opus 4.7 DeepAgents
full search
72.6%
Qwen3.7-Max Claude Code
full search
72.6%
GLM-5.2 Hermes
full search
72.6%
Claude Sonnet 5 DeepAgents
full search
72.6%
GLM-5.1 Claude Code
full
72.4%

Source: The OfficePrism paper

ModelEffortHarnessScoreCost per taskOutput tokens
Kimi K3 full ReAct 76.4%
Kimi K3 search ReAct 76.2%
Kimi K3 full DeepAgents 76%
Kimi K3 search DeepAgents 76%
GLM-5.2 search Claude Code 74.3%
Claude Opus 4.7 search ReAct 74%
Claude Opus 4.8 search ReAct 74%
DeepSeek-V4-Flash full ReAct 73.6%
Claude Sonnet 5 search ReAct 73.1%
DeepSeek-V4-Pro full ReAct 72.9%
Qwen3.7-Plus full Claude Code 72.9%
Kimi K3 search Claude Code 72.9%
Claude Opus 4.7 search DeepAgents 72.6%
Qwen3.7-Max search Claude Code 72.6%
Qwen3.7-Plus search Claude Code 72.6%
GLM-5.2 search Hermes 72.6%
Claude Sonnet 5 search DeepAgents 72.6%
GLM-5.1 search Claude Code 72.4%
Claude Opus 4.8 search Claude Code 72.4%
DeepSeek-V4-Pro search ReAct 72.1%
DeepSeek-V4-Pro search Claude Code 72.1%
GLM-5.2 full ReAct 72.1%
Kimi K3 search OpenClaw 72.1%
Claude Opus 4.7 search Claude Code 71.9%
GPT-5.5 search DeepAgents 71.9%
GLM-5.2 search ReAct 71.9%
Kimi K3 search Hermes 71.9%
GLM-5.1 full Claude Code 71.7%
GLM-5.1 search ReAct 71.4%
GLM-5.2 full Claude Code 71.4%
Qwen 3.6 Max (Preview) search Claude Code 71.2%
Claude Opus 4.8 search DeepAgents 71.2%
Claude Sonnet 5 full DeepAgents 71.2%
GPT-5.5 search ReAct 71%
DeepSeek-V4-Flash search ReAct 71%
Qwen3.7-Plus search DeepAgents 71%
GLM-5.2 search DeepAgents 70.7%
Qwen3.7-Max search DeepAgents 70.2%
GLM-5.2 full DeepAgents 70.2%
GLM-5.2 full Hermes 70.2%
Kimi K3 full Claude Code 70.2%
GLM-5.1 full ReAct 70%
Claude Opus 4.7 full DeepAgents 70%
Claude Opus 4.7 search Hermes 70%
Claude Opus 4.8 full DeepAgents 70%
Qwen3.7-Plus search ReAct 70%
Qwen 3.6 Plus full Claude Code 69.5%
DeepSeek-V4-Flash search Claude Code 69.5%
Claude Opus 4.8 search Hermes 69.5%
Qwen3.7-Plus full ReAct 69.5%
Kimi K3 full Hermes 69.3%
Qwen 3.6 Plus search Claude Code 69%
GLM-5.1 search DeepAgents 69%
Qwen3.7-Max search OpenClaw 69%
Claude Sonnet 5 search Hermes 69%
Claude Sonnet 5 search Claude Code 68.8%
Qwen 3.6 Max (Preview) full Claude Code 68.6%
DeepSeek-V4-Pro full Hermes 68.3%
GPT-5.5 full DeepAgents 68.1%
DeepSeek-V4-Pro search DeepAgents 68.1%
DeepSeek-V4-Pro search Hermes 68.1%
Qwen 3.6 Plus full DeepAgents 67.9%
DeepSeek-V4-Flash full Hermes 67.9%
Qwen3.7-Max full DeepAgents 67.9%
Qwen3.7-Plus full DeepAgents 67.9%
GLM-5.2 full OpenClaw 67.9%
Qwen 3.6 Plus full ReAct 67.6%
DeepSeek-V4-Pro full Claude Code 67.6%
Qwen3.7-Max full Claude Code 67.6%
GLM-5.2 search OpenClaw 67.6%
Qwen 3.6 Max (Preview) search DeepAgents 67.4%
GLM-5.1 full DeepAgents 67.1%
Qwen 3.6 Max (Preview) full ReAct 66.9%
Qwen 3.6 Max (Preview) search Hermes 66.9%
Kimi K3 full OpenClaw 66.7%
DeepSeek-V4-Flash search Hermes 66.4%
DeepSeek-V4-Flash full DeepAgents 66.2%
DeepSeek-V4-Pro full DeepAgents 66.2%
Kimi K2.6 full ReAct 66%
DeepSeek-V4-Pro full OpenClaw 65.7%
GLM-5.1 search OpenClaw 65.5%
DeepSeek-V4-Pro search OpenClaw 65.5%
Qwen3.7-Max full ReAct 65.5%
Qwen 3.6 Max (Preview) search OpenClaw 65.2%
Kimi K2.6 full DeepAgents 65%
Qwen 3.6 Max (Preview) search ReAct 65%
Qwen 3.6 Plus search OpenClaw 64.5%
GLM-5.1 full OpenClaw 64.3%
Qwen3.7-Plus full Hermes 64.3%
GLM-5.1 search Hermes 64%
Qwen 3.6 Plus search ReAct 63.8%
Qwen 3.6 Max (Preview) full DeepAgents 63.6%
Claude Opus 4.7 search OpenClaw 63.3%
Kimi K2.6 search ReAct 63.3%
Qwen3.7-Max search ReAct 63.3%
DeepSeek-V4-Flash full Claude Code 63.1%
GLM-5.1 full Hermes 62.9%
Kimi K2.6 search DeepAgents 62.9%
DeepSeek-V4-Flash search DeepAgents 62.9%
Qwen3.7-Plus search OpenClaw 62.9%
Qwen3.7-Plus search Hermes 62.6%
Kimi K2.6 search Claude Code 62.4%
GPT-5.5 search Claude Code 62.4%
Qwen3.7-Max search Hermes 62.4%
DeepSeek-V4-Flash search OpenClaw 62.1%
DeepSeek-V4-Flash full OpenClaw 61.7%
Kimi K2.6 search Hermes 61.4%
Qwen3.7-Max full OpenClaw 61.4%
Qwen3.7-Max full Hermes 61.4%
Claude Opus 4.8 search OpenClaw 61.4%
Kimi K2.6 full Claude Code 61.2%
Qwen 3.6 Max (Preview) full Hermes 61%
GPT-5.5 search Hermes 60.5%
Kimi K2.6 search OpenClaw 59.8%
Qwen3.7-Plus full OpenClaw 59.8%
Qwen 3.6 Plus full Hermes 59.5%
Qwen 3.6 Plus full OpenClaw 59.3%
Kimi K2.6 full Hermes 59.3%
Qwen 3.6 Plus search DeepAgents 58.3%
Kimi K2.6 full OpenClaw 58.3%
Qwen 3.6 Max (Preview) full OpenClaw 54%
Qwen 3.6 Plus search Hermes 53.6%
GPT-5.5 search OpenClaw 32.9%
Claude Sonnet 5 search OpenClaw 22.6%

Number one over time

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

Moonshot AI
Benchmark paper 1 leader
Kimi K3 76.4%

Source: The OfficePrism 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 Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Kimi K3 Claude Opus 4.7 Qwen 3.6 Plus 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Kimi K3 Qwen 3.6 Plus

Source: Official OfficePrism 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

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (14 entries).
  • Pass Models scored 14 models we track.
  • Pass Newest score The newest is from July 2026, for Kimi K3.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 76.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 124 rows for models we track 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