Benchmark Almanac

Enterprise-Bench L1-L2

Computer, by DevRev + Claude Opus 5 97.9% tops its leaderboard Claude Opus 5 82.9% leads as a model on its own
Not in model cards
Version 1

Agents answer 14 operational questions across connected enterprise records. Tasks require either retrieving facts or combining evidence under business rules.

On November 19, 2026, DevRev Computer with Claude Opus 5 led at 97.86% accuracy across 140 trials. Version 2 has a separate leaderboard.

Made by
DevRev
Released
July 2026
Kind
Apps and MCP servers
Level
University
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text, Actions
Interaction
Agentic
Environment
APIs and tools
Items from
Real work and use
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
34
Entries on its own leaderboard, 10 models we track

Computer, by DevRev + Claude Opus 5 tops the leaderboard with 97.9% Claude Opus 5 leads with 82.9% among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran on its own with several settings, we keep its best. Every run is in the next chart. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. The best model on its own is Claude Opus 5, at 82.9%. Agents built around a model, such as Computer, by DevRev + Claude Opus 5, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

The newest model with a score here came out in September 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 Computer, by DevRev + Claude Opus 5 97.9%
  2. 2 Computer, by DevRev + Claude Opus 5.5 96.4%
  3. 3 Computer, by DevRev + Claude Sonnet 5.5 96.4%
  4. 4 Computer, by DevRev + Claude Opus 4.6 95.7%
  5. 5 Computer, by DevRev + Claude Opus 4.8 94.3%
  6. 6 Computer, by DevRev + GLM-5.2 92.9%
  7. 7 Computer, by DevRev + GLM-5.3 92.9%
  8. 8 Computer, by DevRev + GPT-6 Astra 92.9%
  9. 9 Computer, by DevRev + GPT-6.1 Sol 87.9%
  10. 10 Claude Opus 5 82.9%
  11. 11 Claude Opus 5.5 79.3%
  12. 12 Trinity-EFC + DeepSeek-V4-Pro 78.6%
  13. 13 GPT-6 Astra 75.7%
  14. 14 Grok 4.5 74.3%
  15. 15 GPT-6.1 Sol 74.3%
  16. 16 Opencode + Gemini 4 Argon 73.6%
  17. 17 Opencode + Qwen3.8 Max (0902) 72.9%
  18. 18 OpenCode + GLM-5.3 70.7%
  19. 19 Computer, by DevRev + GPT-5.6 Sol 70%
  20. 20 Claude Sonnet 5.5 65.7%
  21. 21 Opencode + GLM-5.3 65%
  22. 22 Claude Opus 4.8 63.6%
  23. 23 Computer, by DevRev + Gemini 4 Argon 62.1%
  24. 24 Opencode + Mistral Large 4 62.1%
  25. 25 Claude Opus 4.6 61.4%
  26. 26 Claude Code + GLM-5.2 58.6%
  27. 27 Computer, by DevRev + Qwen3.8 Max (0902) 57.9%
  28. 28 Opencode + DeepSeek V4 Pro 0813 56.4%
  29. 29 Hermes + Kimi K3 53.6%
  30. 30 OpenCode + GLM-5.2 52.1%
  31. 31 Claude Code + DeepSeek-V4-Pro 51.4%
  32. 32 GPT-5.6 Sol 51.4%
  33. 33 GPT-6 Luna high 49.3%
  34. 34 OpenCode + Kimi K2.7 Code 32.9%

Source: Official Enterprise-Bench L1-L2 leaderboard

#EntryScoreSettingAgent byReleasedRun onSource
1 Computer, by DevRev + Claude Opus 5 97.9% DevRev Jul 2026 Oct 2026 Official leaderboard
2 Computer, by DevRev + Claude Opus 5.5 96.4% DevRev Sep 2026 Oct 2026 Official leaderboard
3 Computer, by DevRev + Claude Sonnet 5.5 96.4% DevRev Sep 2026 Oct 2026 Official leaderboard
4 Computer, by DevRev + Claude Opus 4.6 95.7% DevRev Feb 2026 Jul 2026 Official leaderboard
5 Computer, by DevRev + Claude Opus 4.8 94.3% DevRev May 2026 Jul 2026 Official leaderboard
6 Computer, by DevRev + GLM-5.2 92.9% DevRev Jun 2026 Aug 2026 Official leaderboard
7 Computer, by DevRev + GLM-5.3 92.9% DevRev Aug 2026 Oct 2026 Official leaderboard
8 Computer, by DevRev + GPT-6 Astra 92.9% DevRev Sep 2026 Oct 2026 Official leaderboard
9 Computer, by DevRev + GPT-6.1 Sol 87.9% DevRev Sep 2026 Oct 2026 Official leaderboard
10 Claude Opus 5 82.9% Claude Code Anthropic Jul 2026 Oct 2026 Official leaderboard
11 Claude Opus 5.5 79.3% Claude Code Anthropic Sep 2026 Oct 2026 Official leaderboard
12 Trinity-EFC + DeepSeek-V4-Pro 78.6% Ajay Raj S Apr 2026 Sep 2026 Official leaderboard
13 GPT-6 Astra 75.7% Codex OpenAI Sep 2026 Oct 2026 Official leaderboard
14 Grok 4.5 74.3% Grok Build xAI Jul 2026 Jul 2026 Official leaderboard
15 GPT-6.1 Sol 74.3% Codex OpenAI Sep 2026 Nov 2026 Official leaderboard
16 Opencode + Gemini 4 Argon 73.6% Anomaly Sep 2026 Oct 2026 Official leaderboard
17 Opencode + Qwen3.8 Max (0902) 72.9% Anomaly Sep 2026 Oct 2026 Official leaderboard
18 OpenCode + GLM-5.3 70.7% Anomaly Aug 2026 Nov 2026 Official leaderboard
19 Computer, by DevRev + GPT-5.6 Sol 70% DevRev Jul 2026 Jul 2026 Official leaderboard
20 Claude Sonnet 5.5 65.7% Claude Code Anthropic Sep 2026 Oct 2026 Official leaderboard
21 Opencode + GLM-5.3 65% Anomaly Aug 2026 Oct 2026 Official leaderboard
22 Claude Opus 4.8 63.6% Claude Code Anthropic May 2026 Jul 2026 Official leaderboard
23 Computer, by DevRev + Gemini 4 Argon 62.1% DevRev Sep 2026 Nov 2026 Official leaderboard
24 Opencode + Mistral Large 4 62.1% Anomaly Oct 2026 Oct 2026 Official leaderboard
25 Claude Opus 4.6 61.4% Claude Code Anthropic Feb 2026 Jul 2026 Official leaderboard
26 Claude Code + GLM-5.2 58.6% Anthropic Jun 2026 Jul 2026 Official leaderboard
27 Computer, by DevRev + Qwen3.8 Max (0902) 57.9% DevRev Sep 2026 Nov 2026 Official leaderboard
28 Opencode + DeepSeek V4 Pro 0813 56.4% Anomaly Aug 2026 Oct 2026 Official leaderboard
29 Hermes + Kimi K3 53.6% Nous Research Jul 2026 Aug 2026 Official leaderboard
30 OpenCode + GLM-5.2 52.1% Anomaly Jun 2026 Jul 2026 Official leaderboard
31 Claude Code + DeepSeek-V4-Pro 51.4% Anthropic Apr 2026 Sep 2026 Official leaderboard
32 GPT-5.6 Sol 51.4% Codex OpenAI Jul 2026 Jul 2026 Official leaderboard
33 GPT-6 Luna 49.3% high, Codex OpenAI Sep 2026 Oct 2026 Official leaderboard
34 OpenCode + Kimi K2.7 Code 32.9% Anomaly Jun 2026 Jul 2026 Official leaderboard
#ModelScoreSettingAgent byReleasedRun onSource
1 Claude Opus 5 82.9% Claude Code Anthropic Jul 2026 Oct 2026 Official leaderboard
2 Claude Opus 5.5 79.3% Claude Code Anthropic Sep 2026 Oct 2026 Official leaderboard
3 GPT-6 Astra 75.7% Codex OpenAI Sep 2026 Oct 2026 Official leaderboard
4 Grok 4.5 74.3% Grok Build xAI Jul 2026 Jul 2026 Official leaderboard
5 GPT-6.1 Sol 74.3% Codex OpenAI Sep 2026 Nov 2026 Official leaderboard
6 Claude Sonnet 5.5 65.7% Claude Code Anthropic Sep 2026 Oct 2026 Official leaderboard
7 Claude Opus 4.8 63.6% Claude Code Anthropic May 2026 Jul 2026 Official leaderboard
8 Claude Opus 4.6 61.4% Claude Code Anthropic Feb 2026 Jul 2026 Official leaderboard
9 GPT-5.6 Sol 51.4% Codex OpenAI Jul 2026 Jul 2026 Official leaderboard
10 GPT-6 Luna 49.3% high, Codex OpenAI Sep 2026 Oct 2026 Official leaderboard

The best score, 97.9%, costs $116 for the full run

Runs on the cost frontier score higher than any run that's cheaper. Computer, by DevRev + GLM-5.3 comes within 5 points of the best for $26 for the full run, 5 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Agents other companies built, in a pale shade of the model’s lab Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $1.00 $10 $100 $1,000 Cost of the full run, log scale Computer, by DevRev + Claude Opus 5 Computer, by DevRev + GLM-5.3 GPT-6.1 Sol Opencode + DeepSeek V4 Pro 0813 GPT-6 Luna, high 0% 20% 40% 60% 80% 100% $1.00 $10 $100 $1,000 Cost of the full run, log scale Computer, by DevRev + Claude Opus 5 GPT-6.1 Sol

Source: Official Enterprise-Bench L1-L2 leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Computer, by DevRev + Claude Opus 5 Computer, by DevRev 97.9% $116 7,765
Computer, by DevRev + Claude Opus 5.5 Computer, by DevRev 96.4% $77 6,527
Computer, by DevRev + Claude Sonnet 5.5 Computer, by DevRev 96.4% $37 5,673
Computer, by DevRev + Claude Opus 4.6 Computer, by DevRev 95.7% $37 3,660
Computer, by DevRev + Claude Opus 4.8 Computer, by DevRev 94.3% $60 4,841
Computer, by DevRev + GLM-5.2 Computer, by DevRev 92.9% 3,651
Computer, by DevRev + GLM-5.3 Computer, by DevRev 92.9% $26 6,741
Computer, by DevRev + GPT-6 Astra Computer, by DevRev 92.9% $213 4,414
Computer, by DevRev + GPT-6.1 Sol Computer, by DevRev 87.9% $40 4,436
Claude Opus 5 Claude Code 82.9% $169 12,652
Claude Opus 5.5 Claude Code 79.3% $63 6,234
Trinity-EFC + DeepSeek-V4-Pro Trinity-EFC 78.6% 3,546
GPT-6 Astra Codex 75.7% $123 1,647
Grok 4.5 Grok Build 74.3% $102 15,455
GPT-6.1 Sol Codex 74.3% $23 3,237
Opencode + Gemini 4 Argon Opencode 73.6% $96 6,669
Opencode + Qwen3.8 Max (0902) Opencode 72.9% $41 3,174
OpenCode + GLM-5.3 OpenCode 70.7% $22 8,292
Computer, by DevRev + GPT-5.6 Sol Computer, by DevRev 70% $335 5,790
Claude Sonnet 5.5 Claude Code 65.7% $25 4,929
Opencode + GLM-5.3 Opencode 65% 9,237
Claude Opus 4.8 Claude Code 63.6% $142 12,045
Computer, by DevRev + Gemini 4 Argon Computer, by DevRev 62.1% $80 10,352
Opencode + Mistral Large 4 Opencode 62.1% $25 3,351
Claude Opus 4.6 Claude Code 61.4% $99 7,544
Claude Code + GLM-5.2 Claude Code 58.6% $88 6,333
Computer, by DevRev + Qwen3.8 Max (0902) Computer, by DevRev 57.9% $51 12,165
Opencode + DeepSeek V4 Pro 0813 Opencode 56.4% $16 4,790
Hermes + Kimi K3 Hermes 53.6% 6,150
OpenCode + GLM-5.2 OpenCode 52.1% 5,022
GPT-5.6 Sol Codex 51.4% $91 4,937
Claude Code + DeepSeek-V4-Pro Claude Code 51.4% 4,018
GPT-6 Luna high Codex 49.3% $1.57 4,366
OpenCode + Kimi K2.7 Code OpenCode 32.9% 4,200

The best score, 82.9%, costs $169 for the full run

Runs on the cost frontier score higher than any run that's cheaper. Claude Opus 5.5 comes within 3.6 points of the best for $63 for the full run, 3 times 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% $1.00 $10 $100 $1,000 Cost of the full run, log scale Claude Opus 5 Claude Opus 5.5 GPT-6.1 Sol GPT-6 Luna, high 0% 20% 40% 60% 80% 100% $1.00 $10 $100 $1,000 Cost of the full run, log scale Claude Opus 5 GPT-6.1 Sol

Source: Official Enterprise-Bench L1-L2 leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 5 Claude Code 82.9% $169 12,652
Claude Opus 5.5 Claude Code 79.3% $63 6,234
GPT-6 Astra Codex 75.7% $123 1,647
Grok 4.5 Grok Build 74.3% $102 15,455
GPT-6.1 Sol Codex 74.3% $23 3,237
Claude Sonnet 5.5 Claude Code 65.7% $25 4,929
Claude Opus 4.8 Claude Code 63.6% $142 12,045
Claude Opus 4.6 Claude Code 61.4% $99 7,544
GPT-5.6 Sol Codex 51.4% $91 4,937
GPT-6 Luna high Codex 49.3% $1.57 4,366

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

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

Anthropic xAI
Own leaderboard Replaced by v2, Nov 2026

Source: Official Enterprise-Bench L1-L2 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 Near the ceiling (85%) Claude Opus 5 Grok 4.5 Claude Opus 4.8 Claude Opus 4.6 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 Grok 4.5

Source: Official Enterprise-Bench L1-L2 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.

Versions

More in Tool use

All 265 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Neon 56.3%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%

Data audit

All 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (34 entries).
  • Pass Models scored 10 models we track, plus 24 other entries on its leaderboard.
  • Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
  • Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 82.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on hub.harborframework.com: 34 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