Benchmark Almanac

RCM-Bench-1

Claude Opus 5 92.7% leads since Jul 2026 Not in model cards

Agents handle healthcare revenue cycle work, such as claim submission, billing and eligibility checks, with tools. Checks cover tool calls, arguments, replies and final system state.

97 scenarios across six workflows, weighted equally. 20 models in the Fireworks index. The maker page renders in the browser only.

Made by Commure Medicine Tasks kept private Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard eng.commure.com
20
Models scored on its own leaderboard

Claude Opus 5 leads with 92.7%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

20 models. Scroll sideways for the rest.

  1. 92.7 Claude Opus 5
  2. 91.5 Gemini 4 Argon
  3. 91.2 Claude Sonnet 5
  4. 91.2 GPT-5.6 Sol
  5. 91.0 Claude Sonnet 5.5
  6. 90.2 Gemini 3.8 Flash
  7. 90.1 Claude Opus 5.5
  8. 89.8 Claude Sonnet 4.6
  9. 89.0 DeepSeek V4.1 Flash
  10. 88.4 Claude Haiku 5.5
  11. 88.3 GLM-5.3
  12. 88.2 Grok 4.7
  13. 87.8 Claude Opus 4.6
  14. 87.4 GPT-6.1 Sol
  15. 87.0 GPT-5.6 Terra
  16. 85.8 GPT-6 Luna
  17. 84.3 GPT-5 mini
  18. 81.5 GPT-6 Sol
  19. 71.5 GPT-5.6 Luna
  20. 70.2 Claude Haiku 4.5
  1. 1 Claude Opus 5 92.7%
  2. 2 Gemini 4 Argon medium 91.5%
  3. 3 Claude Sonnet 5 91.2%
  4. 4 GPT-5.6 Sol medium 91.2%
  5. 5 Claude Sonnet 5.5 91%
  6. 6 Gemini 3.8 Flash medium 90.2%
  7. 7 Claude Opus 5.5 90.1%
  8. 8 Claude Sonnet 4.6 89.8%
  9. 9 DeepSeek V4.1 Flash 89%
  10. 10 Claude Haiku 5.5 88.4%
  11. 11 GLM-5.3 88.3%
  12. 12 Grok 4.7 88.2%
  13. 13 Claude Opus 4.6 87.8%
  14. 14 GPT-6.1 Sol medium 87.4%
  15. 15 GPT-5.6 Terra medium 87%
  16. 16 GPT-6 Luna medium 85.8%
  17. 17 GPT-5 mini 84.3%
  18. 18 GPT-6 Sol medium 81.5%
  19. 19 GPT-5.6 Luna none 71.5%
  20. 20 Claude Haiku 4.5 70.2%
#ModelScoreSettingDateSource
1 Claude Opus 5 92.7% Jul 2026 Official leaderboard
2 Gemini 4 Argon 91.5% medium Sep 2026 Official leaderboard
3 Claude Sonnet 5 91.2% Jun 2026 Official leaderboard
4 GPT-5.6 Sol 91.2% medium Jul 2026 Official leaderboard
5 Claude Sonnet 5.5 91% Sep 2026 Official leaderboard
6 Gemini 3.8 Flash 90.2% medium Sep 2026 Official leaderboard
7 Claude Opus 5.5 90.1% Sep 2026 Official leaderboard
8 Claude Sonnet 4.6 89.8% Feb 2026 Official leaderboard
9 DeepSeek V4.1 Flash 89% Sep 2026 Official leaderboard
10 Claude Haiku 5.5 88.4% Oct 2026 Official leaderboard
11 GLM-5.3 88.3% Aug 2026 Official leaderboard
12 Grok 4.7 88.2% Sep 2026 Official leaderboard
13 Claude Opus 4.6 87.8% Feb 2026 Official leaderboard
14 GPT-6.1 Sol 87.4% medium Sep 2026 Official leaderboard
15 GPT-5.6 Terra 87% medium Jul 2026 Official leaderboard
16 GPT-6 Luna 85.8% medium Sep 2026 Official leaderboard
17 GPT-5 mini 84.3% Aug 2025 Official leaderboard
18 GPT-6 Sol 81.5% medium Sep 2026 Official leaderboard
19 GPT-5.6 Luna 71.5% none Jul 2026 Official leaderboard
20 Claude Haiku 4.5 70.2% Oct 2025 Official leaderboard

The best score, 92.7%, costs $0.15 a task

Runs on the cost frontier score higher than any run that's cheaper. Claude Haiku 5.5 comes within 4.3 points of the best for $0.003 a task, 50 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic Google DeepMind OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 5 Claude Sonnet 5 Gemini 3.8 Flash, medium Claude Haiku 5.5 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 5 GPT-6 Luna, medium
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5 92.7% $0.15
Gemini 4 Argon medium 91.5% $0.09
Claude Sonnet 5 91.2% $0.05
GPT-5.6 Sol medium 91.2% $0.07
Claude Sonnet 5.5 91% $0.05
Gemini 3.8 Flash medium 90.2% $0.02
Claude Opus 5.5 90.1% $0.09
Claude Sonnet 4.6 89.8% $0.05
DeepSeek V4.1 Flash 89%
Claude Haiku 5.5 88.4% $0.003
GLM-5.3 88.3% $0.02
Grok 4.7 88.2% $0.07
Claude Opus 4.6 87.8%
GPT-6.1 Sol medium 87.4% $0.02
GPT-5.6 Terra medium 87% $0.04
GPT-6 Luna medium 85.8% $0.002
GPT-5 mini 84.3% $0.008
GPT-6 Sol medium 81.5% $0.03
GPT-5.6 Luna none 71.5% $0.003
Claude Haiku 4.5 70.2% $0.01

5 models have held number one since 2023

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

OpenAI Anthropic
Own leaderboard Passed 85% in Feb 2026, by Claude Opus 4.6
Claude Opus 5 92.7%

The best score passed 85% in Feb 2026

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

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 Claude Sonnet 4.6 GPT-5 mini 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 GPT-5 mini

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Law, finance and medicine

All 140 in Law, finance and medicine
  • 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%
  • 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 In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items. Solar Pro 4 77.5%
  • 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. Solar Pro 4 93.2%
  • PRBench In use Model answers 1,100 expert-written law and finance questions. LLM judge grades against weighted expert rubrics of 10-30 criteria each. Muse Spark 1.3 61.6%