Benchmark Almanac

Ramp Accounting Bench

Gemini 4 Argon 61.1% leads since November 2026
Not in model cards

Agents reconcile accounts and prepare financial records across 22 simulated companies. A model grades their answers and files against accounting rubrics.

Ramp uses 137 tasks. Mercor’s earlier APEX-Accounting release uses 100 tasks across ten worlds.

Made by
Ramp Labs
Released
November 2026
Kind
Finance
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
Kept private
Measures
Capability
Language
English
Interaction
Agentic
Environment
A simulator
Refresh
One fixed set
Run by
A host
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
18
Models scored on its own leaderboard

Gemini 4 Argon leads with 61.1%

Best score per model 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.

  1. 1 Gemini 4 Argon 61.1%
  2. 2 Gemini 4 Carbon 56.5%
  3. 3 Claude Fable 5.2 52.2%
  4. 4 Gemini 4 Hydrogen 51.7%
  5. 5 Claude Opus 5.5 50.5%
  6. 6 GPT-6.2 Astra 50.4%
  7. 7 Muse Wonder 48.7%
  8. 8 GPT-6.2 Sol 45.6%
  9. 9 Grok 4.7 44.1%
  10. 10 GLM-5.3 43.1%
  11. 11 Muse Spark 1.3 42.5%
  12. 12 Claude Sonnet 5.5 42.5%
  13. 13 Claude Haiku 5.5 42.3%
  14. 14 Qwen 3.8 Max 41.2%
  15. 15 DeepSeek V4.1 Flash 40.8%
  16. 16 Kimi K3 39.2%
  17. 17 GLM-5.3-Flash 35.7%
  18. 18 GPT-6 Luna 34%

Source: Official Ramp Accounting Bench leaderboard

#ModelScoreSettingReleasedRun onSource
1 Gemini 4 Argon 61.1% Mercor Archipelago Sep 2026 Oct 2026 Official leaderboard
2 Gemini 4 Carbon 56.5% Mercor Archipelago Oct 2026 Nov 2026 Official leaderboard
3 Claude Fable 5.2 52.2% Mercor Archipelago Nov 2026 Nov 2026 Official leaderboard
4 Gemini 4 Hydrogen 51.7% Mercor Archipelago Nov 2026 Nov 2026 Official leaderboard
5 Claude Opus 5.5 50.5% Mercor Archipelago Sep 2026 Sep 2026 Official leaderboard
6 GPT-6.2 Astra 50.4% Mercor Archipelago Nov 2026 Oct 2026 Official leaderboard
7 Muse Wonder 48.7% Mercor Archipelago Oct 2026 Oct 2026 Official leaderboard
8 GPT-6.2 Sol 45.6% Mercor Archipelago Nov 2026 Nov 2026 Official leaderboard
9 Grok 4.7 44.1% Mercor Archipelago Sep 2026 Sep 2026 Official leaderboard
10 GLM-5.3 43.1% Mercor Archipelago Aug 2026 Sep 2026 Official leaderboard
11 Muse Spark 1.3 42.5% Mercor Archipelago Sep 2026 Oct 2026 Official leaderboard
12 Claude Sonnet 5.5 42.5% Mercor Archipelago Sep 2026 Sep 2026 Official leaderboard
13 Claude Haiku 5.5 42.3% Mercor Archipelago Oct 2026 Oct 2026 Official leaderboard
14 Qwen 3.8 Max 41.2% Mercor Archipelago Aug 2026 Aug 2026 Official leaderboard
15 DeepSeek V4.1 Flash 40.8% Mercor Archipelago Sep 2026 Sep 2026 Official leaderboard
16 Kimi K3 39.2% Mercor Archipelago Jul 2026 Aug 2026 Official leaderboard
17 GLM-5.3-Flash 35.7% Mercor Archipelago Aug 2026 Aug 2026 Official leaderboard
18 GPT-6 Luna 34% Mercor Archipelago Sep 2026 Sep 2026 Official leaderboard

The best score, 61.1%, costs $3.03 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.

DeepSeek Anthropic OpenAI Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Argon Gemini 4 Hydrogen GPT-6.2 Sol Claude Haiku 5.5 GPT-6 Luna 0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Argon Gemini 4 Hydrogen

Source: Official Ramp Accounting Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Argon Mercor Archipelago 61.1% $3.03 86,405
Gemini 4 Carbon Mercor Archipelago 56.5% $4.12 110,538
Claude Fable 5.2 Mercor Archipelago 52.2% $3.71 34,457
Gemini 4 Hydrogen Mercor Archipelago 51.7% $1.81 65,182
Claude Opus 5.5 Mercor Archipelago 50.5% $1.34 20,064
GPT-6.2 Astra Mercor Archipelago 50.4% $1.31 8,342
Muse Wonder Mercor Archipelago 48.7% $2.65 23,861
GPT-6.2 Sol Mercor Archipelago 45.6% $0.41 9,736
Grok 4.7 Mercor Archipelago 44.1% $3.11 55,106
GLM-5.3 Mercor Archipelago 43.1% $2.16 139,788
Muse Spark 1.3 Mercor Archipelago 42.5% $1.56 30,688
Claude Sonnet 5.5 Mercor Archipelago 42.5% $0.67 20,058
Claude Haiku 5.5 Mercor Archipelago 42.3% $0.21 52,096
Qwen 3.8 Max Mercor Archipelago 41.2% $2.85 63,014
DeepSeek V4.1 Flash Mercor Archipelago 40.8% $0.12 62,292
Kimi K3 Mercor Archipelago 39.2% $1.55 30,887
GLM-5.3-Flash Mercor Archipelago 35.7% $0.18 111,748
GPT-6 Luna Mercor Archipelago 34% $0.05 29,709

Number one over time

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

Google DeepMind
Own leaderboard 1 leader

Source: Official Ramp Accounting Bench 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% 2027 Near the ceiling (85%) Gemini 4 Argon Claude Opus 5.5 Grok 4.7 Qwen 3.8 Max 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Claude Opus 5.5

Source: Official Ramp Accounting 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 Law, finance and medicine

All 637 in Law, finance and medicine
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy. Ling-3.1-flash 81.7%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. o1-preview 92.7%
  • Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%

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 (18 entries).
  • Pass Models scored 18 models we track.
  • Pass Newest score The newest is from November 2026, for GPT-6.2 Astra.
  • Pass Frontier models 6 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Pass Headroom The best is 61.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on labs.ramp.com: 18 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