Benchmark Almanac

InsureBench Multi-functional LLMs

Claude Opus 4.8 96.3% leads since June 2026
Not in model cards

Models read insurance files and answer questions about claims, coverage, and actuarial decisions. InsureThing combines automatic answer checks with rubric grades for judgment tasks.

The report contains 14 model and harness combinations. Task coverage varies by combination, and the site publishes three sample tasks.

Made by
InsureThing
Released
June 2026
Kind
Finance
Level
University
Scored
Graded by a model
Answers
Agent task
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Text, Documents → Text
Interaction
Single turn
Refresh
One fixed set
Run by
A host
In views
Law and finance
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
June 2026
Scores refreshed
November 20th, 2026
14
Models scored on its own leaderboard

Claude Opus 4.8 leads with 96.3% among models released by June 2026

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.

The newest model with a score here came out in June 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 Claude Opus 4.8 96.3%
  2. 2 GPT-5.5 95.4%
  3. 3 Nemotron 3 Ultra 95.1%
  4. 4 Gemini 3.5 Flash low 94.8%
  5. 5 Tencent Hy3 preview 94.1%
  6. 6 MiMo-V2.5 94%
  7. 7 Claude Haiku 4.5 93.6%
  8. 8 Nemotron 3 Super 92.5%
  9. 9 Gemini 3.1 Pro high 91.6%
  10. 10 GPT-5.4 Mini 88.8%
  11. 11 gpt-oss-120b 87.6%
  12. 12 DeepSeek-V4-Flash 83.1%
  13. 13 Gemma 4 26B A4B 76.1%
  14. 14 Gemini 2.5 Flash-Lite Sep 2025 67.3%

Source: Official InsureBench Multi-functional LLMs leaderboard

#ModelScore ±SettingDateSource
1 Claude Opus 4.8 96.3% 93.1 to 99 claude_code May 2026 Official leaderboard
2 GPT-5.5 95.4% 92.6 to 97.9 codex_plain Apr 2026 Official leaderboard
3 Nemotron 3 Ultra 95.1% 92.3 to 97.6 tooled_reader_v3 Jun 2026 Official leaderboard
4 Gemini 3.5 Flash 94.8% 91.5 to 97.4 low, antigravity May 2026 Official leaderboard
5 Tencent Hy3 preview 94.1% 91.1 to 96.5 tooled_reader_v3 Apr 2026 Official leaderboard
6 MiMo-V2.5 94% 90 to 97.2 tooled_reader_v3 Apr 2026 Official leaderboard
7 Claude Haiku 4.5 93.6% 89.7 to 96.9 claude_code Oct 2025 Official leaderboard
8 Nemotron 3 Super 92.5% ±3.6 tooled_reader_v3 Mar 2026 Official leaderboard
9 Gemini 3.1 Pro 91.6% 87.6 to 96 high, antigravity Feb 2026 Official leaderboard
10 GPT-5.4 Mini 88.8% 84 to 93.4 codex_plain Mar 2026 Official leaderboard
11 gpt-oss-120b 87.6% ±4.1 tooled_reader_v3 Aug 2025 Official leaderboard
12 DeepSeek-V4-Flash 83.1% ±4.3 tooled_reader_v3 Apr 2026 Official leaderboard
13 Gemma 4 26B A4B 76.1% ±5.9 tooled_reader_v3 Apr 2026 Official leaderboard
14 Gemini 2.5 Flash-Lite Sep 2025 67.3% ±6.7 tooled_reader_v3 Sep 2025 Official leaderboard

The best score, 96.3%, costs $0.40 a task

Runs on the cost frontier score higher than any run that's cheaper. Nemotron 3 Super comes within 3.8 points of the best for $0.0008 a task, 490 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI NVIDIA Google DeepMind Xiaomi Corp Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 4.8 GPT-5.5 Nemotron 3 Ultra gpt-oss-120b Gemma 4 26B A4B 0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 4.8 MiMo-V2.5

Source: Official InsureBench Multi-functional LLMs leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 4.8 claude_code 96.3% $0.40
GPT-5.5 codex_plain 95.4% $0.08
Nemotron 3 Ultra tooled_reader_v3 95.1% $0.002
Gemini 3.5 Flash low antigravity 94.8% $0.005
Tencent Hy3 preview tooled_reader_v3 94.1% $0.002
MiMo-V2.5 tooled_reader_v3 94% $0.001
Claude Haiku 4.5 claude_code 93.6% $0.03
Nemotron 3 Super tooled_reader_v3 92.5% $0.0008
Gemini 3.1 Pro high antigravity 91.6% $0.05
GPT-5.4 Mini codex_plain 88.8% $0.03
gpt-oss-120b tooled_reader_v3 87.6% $0.0005
DeepSeek-V4-Flash tooled_reader_v3 83.1% $0.0007
Gemma 4 26B A4B tooled_reader_v3 76.1% $0.0005
Gemini 2.5 Flash-Lite Sep 2025 tooled_reader_v3 67.3% $0.001

Number one over time

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

Anthropic
Own leaderboard Passed 85% in Jun 2026, by Claude Opus 4.8

Source: Official InsureBench Multi-functional LLMs leaderboard

Gpt-oss-120b on the official leaderboard passed 85% in August 2025

Models released before it came out already scored that high. 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 4.8 Claude Haiku 4.5 gpt-oss-120b 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 4.8 Claude Haiku 4.5

Source: Official InsureBench Multi-functional LLMs 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 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (14 entries).
  • Pass Models scored 14 models we track.
  • Pass Newest score The newest is from June 2026, for Nemotron 3 Ultra.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5.5 (#21).
  • Note Headroom Saturated. The best is 96.3%, past 85% since August 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on insurebench.insure-thing.com: 14 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