Benchmark Almanac

DataBench v1.1

Claude Opus 5.5 70.5% leads since September 2026
Not in model cards

Agents answer ambiguous business questions and make recommendations from a synthetic company warehouse. Model judges assess the reasoning, supporting evidence, artifacts, and final recommendation against task rubrics.

Version 1 launched with 100 tasks in August 2026. Version 1.1 changed judges and rubrics on September 17 to penalize misleading answers. The November additions use one repetition per model configuration. This is unrelated to Cardiff NLP DataBench.

Made by
Hex
Released
September 2026
Kind
Running a business
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
Kept private
Measures
Capability
Language
English
In and out
Text, Tables → Text, Actions, Tables
Interaction
Agentic
Environment
APIs and tools
Items from
Generated
Run by
A host
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
22
Entries on its own leaderboard, 21 models we track

Claude Opus 5.5 leads with 70.5% Claude Opus 5.5 leads with 70.5%

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. Every run is in the next chart. 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.

  1. 1 Claude Opus 5.5 Max 70.5%
  2. 2 GPT-6 Sol XHigh 61.3%
  3. 3 GPT-6.1 Sol Medium 58%
  4. 4 GPT-6 Astra Low 55%
  5. 5 GPT-6 Luna Max 52.3%
  6. 6 GPT-6.1 Terra High 51%
  7. 7 Kimi K3.1 Pro Max 48%
  8. 8 GPT-5.6 Sol XHigh 40.3%
  9. 9 GPT-5.6 Terra XHigh 39.3%
  10. 10 GPT-5.6 Luna XHigh 35.3%
  11. 11 Grok 4.6 XHigh 28%
  12. 12 Claude Haiku 5.5 Medium 28%
  13. 13 Claude Fable 5.1 Max 27%
  14. 14 Claude Sonnet 5.5 Medium 22%
  15. 15 Claude Opus 5 Max 19.7%
  16. 16 Claude Opus 4.8 XHigh 18.7%
  17. 17 Claude Fable 5 Max 18%
  18. 18 GLM-5.3 Max 16.3%
  19. 19 Claude Sonnet 5 XHigh 14.3%
  20. 20 GLM-5.2 High 13.3%
  21. 21 DeepSeek V4.1 Max 13%
  22. 22 Kimi K2.7 Code Medium 12.7%

Source: Official DataBench v1.1 leaderboard

#EntryScoreSettingDateSource
1 Claude Opus 5.5 70.5% Max Sep 2026 Official leaderboard
2 GPT-6 Sol 61.3% XHigh Sep 2026 Official leaderboard
3 GPT-6.1 Sol 58% Medium Sep 2026 Official leaderboard
4 GPT-6 Astra 55% Low Sep 2026 Official leaderboard
5 GPT-6 Luna 52.3% Max Sep 2026 Official leaderboard
6 GPT-6.1 Terra 51% High Oct 2026 Official leaderboard
7 Kimi K3.1 Pro 48% Max Oct 2026 Official leaderboard
8 GPT-5.6 Sol 40.3% XHigh Jul 2026 Official leaderboard
9 GPT-5.6 Terra 39.3% XHigh Jul 2026 Official leaderboard
10 GPT-5.6 Luna 35.3% XHigh Jul 2026 Official leaderboard
11 Grok 4.6 28% XHigh Aug 2026 Official leaderboard
12 Claude Haiku 5.5 28% Medium Oct 2026 Official leaderboard
13 Claude Fable 5.1 27% Max Sep 2026 Official leaderboard
14 Claude Sonnet 5.5 22% Medium Sep 2026 Official leaderboard
15 Claude Opus 5 19.7% Max Jul 2026 Official leaderboard
16 Claude Opus 4.8 18.7% XHigh May 2026 Official leaderboard
17 Claude Fable 5 18% Max Jun 2026 Official leaderboard
18 GLM-5.3 16.3% Max Aug 2026 Official leaderboard
19 Claude Sonnet 5 14.3% XHigh Jun 2026 Official leaderboard
20 GLM-5.2 13.3% High Jun 2026 Official leaderboard
21 DeepSeek V4.1 13% Max Official leaderboard
22 Kimi K2.7 Code 12.7% Medium Jun 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Opus 5.5 70.5% Max Sep 2026 Official leaderboard
2 GPT-6 Sol 61.3% XHigh Sep 2026 Official leaderboard
3 GPT-6.1 Sol 58% Medium Sep 2026 Official leaderboard
4 GPT-6 Astra 55% Low Sep 2026 Official leaderboard
5 GPT-6 Luna 52.3% Max Sep 2026 Official leaderboard
6 GPT-6.1 Terra 51% High Oct 2026 Official leaderboard
7 Kimi K3.1 Pro 48% Max Oct 2026 Official leaderboard
8 GPT-5.6 Sol 40.3% XHigh Jul 2026 Official leaderboard
9 GPT-5.6 Terra 39.3% XHigh Jul 2026 Official leaderboard
10 GPT-5.6 Luna 35.3% XHigh Jul 2026 Official leaderboard
11 Grok 4.6 28% XHigh Aug 2026 Official leaderboard
12 Claude Haiku 5.5 28% Medium Oct 2026 Official leaderboard
13 Claude Fable 5.1 27% Max Sep 2026 Official leaderboard
14 Claude Sonnet 5.5 22% Medium Sep 2026 Official leaderboard
15 Claude Opus 5 19.7% Max Jul 2026 Official leaderboard
16 Claude Opus 4.8 18.7% XHigh May 2026 Official leaderboard
17 Claude Fable 5 18% Max Jun 2026 Official leaderboard
18 GLM-5.3 16.3% Max Aug 2026 Official leaderboard
19 Claude Sonnet 5 14.3% XHigh Jun 2026 Official leaderboard
20 GLM-5.2 13.3% High Jun 2026 Official leaderboard
21 Kimi K2.7 Code 12.7% Medium Jun 2026 Official leaderboard

The best score, 70.5%, costs $3.57 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Opus 5.5, Max Claude Opus 5.5, High GPT-6.1 Sol, Medium GPT-6 Luna, Max GPT-6 Luna, High GPT-6 Luna, Low 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Opus 5.5, Max Claude Opus 5.5, High

Source: Official DataBench v1.1 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 Max 70.5% $3.57 115,101
Claude Opus 5.5 XHigh 67% $2.16 62,782
Claude Opus 5.5 High 65% $1.25 30,534
Claude Opus 5.5 Medium 62.5% $0.97 22,482
GPT-6 Sol XHigh 61.3% $0.58 14,596
GPT-6.1 Sol Medium 58% $0.25 4,350
GPT-6 Sol Max 57% $0.98 29,927
GPT-6.1 Sol Max 57% $0.50 16,306
GPT-6 Astra Low 55% $0.85 2,737
GPT-6 Astra Max 55% $2.11 14,828
GPT-6.1 Sol High 55% $0.38 7,747
GPT-6.1 Sol XHigh 55% $0.44 10,024
GPT-6 Sol High 54.7% $0.39 8,465
Claude Opus 5.5 Low 54% $0.61 13,011
GPT-6 Astra Medium 52.7% $1.24 4,855
GPT-6 Astra XHigh 52.3% $1.75 9,293
GPT-6 Luna Max 52.3% $0.06 37,928
GPT-6 Astra High 52% $1.61 7,696
GPT-6 Sol Medium 51.3% $0.27 4,775
GPT-6.1 Terra High 51% $0.19 13,313
GPT-6 Sol Low 50.3% $0.17 2,752
GPT-6.1 Terra XHigh 50% $0.27 19,298
GPT-6.1 Terra Max 49% $0.45 35,458
GPT-6.1 Sol Low 48% $0.21 3,112
Kimi K3.1 Pro Max 48% $0.49 20,747
GPT-6 Luna High 46% $0.03 15,545
GPT-6 Luna XHigh 46% $0.04 19,322
GPT-6.1 Terra Low 46% $0.10 7,680
GPT-6.1 Terra Medium 44% $0.15 10,679
GPT-6 Luna Medium 42.7% $0.02 9,118
Kimi K3.1 Pro High 41% $0.33 13,044
GPT-5.6 Sol XHigh 40.3% $0.98 11,427
GPT-5.6 Sol Medium 39.7% $0.61 5,560
GPT-5.6 Terra XHigh 39.3% $0.30 9,486
GPT-5.6 Sol High 38% $0.75 7,967
GPT-5.6 Luna XHigh 35.3% $0.04 11,003
GPT-5.6 Terra High 32.7% $0.20 5,378
GPT-5.6 Luna High 31.3% $0.03 7,884
GPT-6 Luna Low 30.3% $0.01 2,619
GPT-5.6 Sol Low 29.7% $0.37 2,809
GPT-5.6 Terra Medium 29.3% $0.15 3,165
Kimi K3.1 Pro Low 29% $0.23 7,969
Grok 4.6 XHigh 28% $0.96 23,332
Claude Haiku 5.5 Medium 28% $0.03 16,229
Claude Haiku 5.5 High 28% $0.06 28,795
GPT-5.6 Terra Low 27% $0.14 2,821
Grok 4.6 Low 27% $0.94 24,637
Grok 4.6 High 27% $0.93 23,492
Claude Fable 5.1 Max 27% $3.36 31,332
Grok 4.6 Medium 26% $0.97 25,403
GPT-5.6 Luna Medium 24% $0.02 3,838
Claude Haiku 5.5 Low 23% $0.02 10,368
Claude Sonnet 5.5 Medium 22% $0.19 3,563
Claude Fable 5.1 XHigh 21.7% $2.50 21,083
Claude Fable 5.1 Medium 21% $1.18 8,117
Claude Fable 5.1 High 20.3% $1.57 11,420
Claude Opus 5 Max 19.7% $2.11 29,091
GPT-5.6 Luna Low 19% $0.01 2,349
Claude Opus 4.8 XHigh 18.7% $1.20 20,039
Claude Fable 5 Max 18% $3.49 27,499
Claude Opus 5 High 18% $1.61 18,898
Claude Opus 5 XHigh 17.3% $1.89 24,172
Claude Opus 4.8 Medium 17% $0.76 10,074
Claude Opus 5 Medium 17% $1.10 11,427
GLM-5.3 Max 16.3% $0.64 38,848
Claude Fable 5.1 Low 16.3% $0.92 6,069
Claude Fable 5 High 16% $1.88 10,213
Claude Opus 5 Low 15.7% $0.74 6,923
Claude Opus 4.8 Max 15% $1.78 34,705
Claude Fable 5 Low 15% $1.14 4,707
Claude Fable 5 Medium 14.7% $1.50 7,238
GLM-5.3 High 14.7% $0.36 14,875
Claude Sonnet 5 XHigh 14.3% $0.62 20,165
Claude Opus 4.8 High 14% $0.86 12,219
Claude Sonnet 5 High 13.7% $0.49 14,321
GLM-5.2 High 13.3% $0.25 10,494
Claude Sonnet 5 Medium 13.3% $0.36 9,081
Kimi K2.7 Code Medium 12.7% $0.28 11,769
Claude Sonnet 5 Max 12.3% $1.07 43,089
Claude Opus 4.8 Low 12% $0.60 6,890
Claude Sonnet 5 Low 11.7% $0.26 5,866
GLM-5.2 Max 9.00% $0.32 19,073
GLM-5.3 Low 8.70% $0.20 5,341

The best score, 70.5%, costs $3.57 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Opus 5.5, Max Claude Opus 5.5, High GPT-6.1 Sol, Medium GPT-6 Luna, Max GPT-6 Luna, High GPT-6 Luna, Low 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Claude Opus 5.5, Max Claude Opus 5.5, High

Source: Official DataBench v1.1 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 Max 70.5% $3.57 115,101
Claude Opus 5.5 XHigh 67% $2.16 62,782
Claude Opus 5.5 High 65% $1.25 30,534
Claude Opus 5.5 Medium 62.5% $0.97 22,482
GPT-6 Sol XHigh 61.3% $0.58 14,596
GPT-6.1 Sol Medium 58% $0.25 4,350
GPT-6 Sol Max 57% $0.98 29,927
GPT-6.1 Sol Max 57% $0.50 16,306
GPT-6 Astra Low 55% $0.85 2,737
GPT-6 Astra Max 55% $2.11 14,828
GPT-6.1 Sol High 55% $0.38 7,747
GPT-6.1 Sol XHigh 55% $0.44 10,024
GPT-6 Sol High 54.7% $0.39 8,465
Claude Opus 5.5 Low 54% $0.61 13,011
GPT-6 Astra Medium 52.7% $1.24 4,855
GPT-6 Astra XHigh 52.3% $1.75 9,293
GPT-6 Luna Max 52.3% $0.06 37,928
GPT-6 Astra High 52% $1.61 7,696
GPT-6 Sol Medium 51.3% $0.27 4,775
GPT-6.1 Terra High 51% $0.19 13,313
GPT-6 Sol Low 50.3% $0.17 2,752
GPT-6.1 Terra XHigh 50% $0.27 19,298
GPT-6.1 Terra Max 49% $0.45 35,458
GPT-6.1 Sol Low 48% $0.21 3,112
Kimi K3.1 Pro Max 48% $0.49 20,747
GPT-6 Luna High 46% $0.03 15,545
GPT-6 Luna XHigh 46% $0.04 19,322
GPT-6.1 Terra Low 46% $0.10 7,680
GPT-6.1 Terra Medium 44% $0.15 10,679
GPT-6 Luna Medium 42.7% $0.02 9,118
Kimi K3.1 Pro High 41% $0.33 13,044
GPT-5.6 Sol XHigh 40.3% $0.98 11,427
GPT-5.6 Sol Medium 39.7% $0.61 5,560
GPT-5.6 Terra XHigh 39.3% $0.30 9,486
GPT-5.6 Sol High 38% $0.75 7,967
GPT-5.6 Luna XHigh 35.3% $0.04 11,003
GPT-5.6 Terra High 32.7% $0.20 5,378
GPT-5.6 Luna High 31.3% $0.03 7,884
GPT-6 Luna Low 30.3% $0.01 2,619
GPT-5.6 Sol Low 29.7% $0.37 2,809
GPT-5.6 Terra Medium 29.3% $0.15 3,165
Kimi K3.1 Pro Low 29% $0.23 7,969
Grok 4.6 XHigh 28% $0.96 23,332
Claude Haiku 5.5 Medium 28% $0.03 16,229
Claude Haiku 5.5 High 28% $0.06 28,795
GPT-5.6 Terra Low 27% $0.14 2,821
Grok 4.6 Low 27% $0.94 24,637
Grok 4.6 High 27% $0.93 23,492
Claude Fable 5.1 Max 27% $3.36 31,332
Grok 4.6 Medium 26% $0.97 25,403
GPT-5.6 Luna Medium 24% $0.02 3,838
Claude Haiku 5.5 Low 23% $0.02 10,368
Claude Sonnet 5.5 Medium 22% $0.19 3,563
Claude Fable 5.1 XHigh 21.7% $2.50 21,083
Claude Fable 5.1 Medium 21% $1.18 8,117
Claude Fable 5.1 High 20.3% $1.57 11,420
Claude Opus 5 Max 19.7% $2.11 29,091
GPT-5.6 Luna Low 19% $0.01 2,349
Claude Opus 4.8 XHigh 18.7% $1.20 20,039
Claude Fable 5 Max 18% $3.49 27,499
Claude Opus 5 High 18% $1.61 18,898
Claude Opus 5 XHigh 17.3% $1.89 24,172
Claude Opus 4.8 Medium 17% $0.76 10,074
Claude Opus 5 Medium 17% $1.10 11,427
GLM-5.3 Max 16.3% $0.64 38,848
Claude Fable 5.1 Low 16.3% $0.92 6,069
Claude Fable 5 High 16% $1.88 10,213
Claude Opus 5 Low 15.7% $0.74 6,923
Claude Opus 4.8 Max 15% $1.78 34,705
Claude Fable 5 Low 15% $1.14 4,707
Claude Fable 5 Medium 14.7% $1.50 7,238
GLM-5.3 High 14.7% $0.36 14,875
Claude Sonnet 5 XHigh 14.3% $0.62 20,165
Claude Opus 4.8 High 14% $0.86 12,219
Claude Sonnet 5 High 13.7% $0.49 14,321
GLM-5.2 High 13.3% $0.25 10,494
Claude Sonnet 5 Medium 13.3% $0.36 9,081
Kimi K2.7 Code Medium 12.7% $0.28 11,769
Claude Sonnet 5 Max 12.3% $1.07 43,089
Claude Opus 4.8 Low 12% $0.60 6,890
Claude Sonnet 5 Low 11.7% $0.26 5,866
GLM-5.2 Max 9.00% $0.32 19,073
GLM-5.3 Low 8.70% $0.20 5,341

3 models have held number one on the official leaderboard since its release in September 2026

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

OpenAI Anthropic
Own leaderboard 3 leaders

Source: Official DataBench v1.1 leaderboard

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
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5.5 GPT-6 Astra GPT-5.6 Sol Claude Opus 4.8 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5.5 GPT-6 Astra

Source: Official DataBench v1.1 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 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (22 entries).
  • Pass Models scored 21 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from October 2026, for GPT-6.1 Terra.
  • Pass Frontier models 6 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 70.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on hex.tech: 83 rows for models we track and 1 other entry 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