Benchmark Almanac

WearableQA

Gemini 3.1 Pro 72.9% leads since September 2026
Not in model cards

Models answer 4,084 ten-option questions using longitudinal wearable measurements, blood biomarkers, and demographics from 200 users. Questions test computations and physiological interpretation across single or multiple signals. Accuracy compares the selected option with the answer key.

Records contain missing days and device noise. Time-series serialization changes results, so comparisons need the same representation. The released rendered prompts can be long.

Released
September 2026
Kind
Medicine
Level
Expert
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Time series, Tables → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
April 2026
Scores refreshed
November 19th, 2026

In the authors’ words

Recent advances in wearable sensing enable continuous monitoring of physiological and behavioral signals, yet existing benchmarks rarely evaluate whether AI systems can reason over a real user's longitudinal wearable record. We introduce WearableQA, a benchmark comprising 4,084 10-option multiple-choice questions constructed from the wearable time series, blood biomarkers, and demographics of 200 real users, each with up to 500 days of daily measurements. WearableQA preserves authentic wearable distributions that include device noise and inter-individual variability. To evaluate distinct reasoning capabilities, we introduce 16 question types organized along two complementary axes: data versus health reasoning, which distinguishes computation over longitudinal measurements from physiological interpretation; and single- versus cross-signal reasoning, which separates reasoning about individual signals from the integration of multiple signals. To construct reliable questions at scale, we adopt a dual-grounding framework that combines literature-grounded physiological findings with statistically validated population-grounded physiological patterns. This enables the capture of meaningful relationships observed in real-world wearable data. Evaluation of 14 proprietary and open-source LLMs demonstrates that WearableQA effectively differentiates model capabilities, with performance ranging from 19.6% to 72.9% against a 10% chance baseline. Moreover, WearableQA remains far from solved: most models achieve accuracies below 60%. Overall, WearableQA provides a realistic and diagnostic benchmark for evaluating LLM reasoning over real-world wearable data.

Abstract of WearableQA: A Benchmark for Health Reasoning over Real-World Wearable Data (arXiv:2609.05405, CC0)
14
Entries in its paper, 12 models we track

Gemini 3.1 Pro leads with 72.9% among models released by April 2026 Gemini 3.1 Pro leads with 72.9% among models released by April 2026

Best score per entry from the benchmark's paper, read on November 19th, 2026. A paper’s table doesn’t add newer models. 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.

The newest model with a score here came out in April 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 Gemini 3.1 Pro row input; CoT 72.9%
  2. 2 Claude Opus 4.6 row input; CoT 60.2%
  3. 3 Claude Sonnet 4 row input; CoT 53.2%
  4. 4 GPT-5.4 row input; CoT 51.2%
  5. 5 Gemini-2.5-Pro (snapshot unspecified) row input; CoT 50.6%
  6. 6 Gemma 4 26B A4B row input; CoT 42.5%
  7. 7 GPT-4o (snapshot unspecified) row input; CoT 34.7%
  8. 8 Mistral Small 3.1 row input; CoT 29.3%
  9. 9 Llama 3.3 70B row input; CoT 29.2%
  10. 10 Gemma 3 27B row input; CoT 26.2%
  11. 11 Gemma 3 12B row input; CoT 24.6%
  12. 12 Llama 3.1-8B row input; CoT 23.8%
  13. 13 Gemma 3 4B row input; direct answer 20.6%
  14. 14 Llama 3.2 3B row input; CoT 19.6%

Source: The WearableQA paper

#EntryScoreSettingDateSource
1 Gemini 3.1 Pro 72.9% row input; CoT Feb 2026 Official leaderboard
2 Claude Opus 4.6 60.2% row input; CoT Feb 2026 Official leaderboard
3 Claude Sonnet 4 53.2% row input; CoT May 2025 Official leaderboard
4 GPT-5.4 51.2% row input; CoT Mar 2026 Official leaderboard
5 Gemini-2.5-Pro (snapshot unspecified) 50.6% row input; CoT Official leaderboard
6 Gemma 4 26B A4B 42.5% row input; CoT Apr 2026 Official leaderboard
7 GPT-4o (snapshot unspecified) 34.7% row input; CoT Official leaderboard
8 Mistral Small 3.1 29.3% row input; CoT Mar 2025 Official leaderboard
9 Llama 3.3 70B 29.2% row input; CoT Dec 2024 Official leaderboard
10 Gemma 3 27B 26.2% row input; CoT Mar 2025 Official leaderboard
11 Gemma 3 12B 24.6% row input; CoT Mar 2025 Official leaderboard
12 Llama 3.1-8B 23.8% row input; CoT Jul 2024 Official leaderboard
13 Gemma 3 4B 20.6% row input; direct answer Mar 2025 Official leaderboard
14 Llama 3.2 3B 19.6% row input; CoT Sep 2024 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 72.9% row input; CoT Feb 2026 Official leaderboard
2 Claude Opus 4.6 60.2% row input; CoT Feb 2026 Official leaderboard
3 Claude Sonnet 4 53.2% row input; CoT May 2025 Official leaderboard
4 GPT-5.4 51.2% row input; CoT Mar 2026 Official leaderboard
5 Gemma 4 26B A4B 42.5% row input; CoT Apr 2026 Official leaderboard
6 Mistral Small 3.1 29.3% row input; CoT Mar 2025 Official leaderboard
7 Llama 3.3 70B 29.2% row input; CoT Dec 2024 Official leaderboard
8 Gemma 3 27B 26.2% row input; CoT Mar 2025 Official leaderboard
9 Gemma 3 12B 24.6% row input; CoT Mar 2025 Official leaderboard
10 Llama 3.1-8B 23.8% row input; CoT Jul 2024 Official leaderboard
11 Gemma 3 4B 20.6% row input; direct answer Mar 2025 Official leaderboard
12 Llama 3.2 3B 19.6% row input; CoT Sep 2024 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3.1 Pro
row input; …
72.9%
Claude Opus 4.6
row input; … row input; …
60.2%
Claude Sonnet 4
row input; …
53.2%
GPT-5.4
row input; … row input; …
51.2%
Gemma 4 26B A4B
row input; … row input; …
42.5%
Mistral Small 3.1
row input; …
29.3%
Llama 3.3 70B
row input; …
29.2%
Gemma 3 27B
row input; …
26.2%
Gemma 3 12B
row input; …
24.6%
Llama 3.1-8B
row input; …
23.8%
Gemma 3 4B
row input; …
20.6%
Llama 3.2 3B
row input; …
19.6%

Source: The WearableQA paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro row input; CoT 72.9%
Gemini 3.1 Pro row input; direct answer 71.3%
Claude Opus 4.6 row input; CoT 60.2%
Claude Sonnet 4 row input; CoT 53.2%
GPT-5.4 row input; CoT 51.2%
Claude Sonnet 4 row input; direct answer 48.3%
Gemma 4 26B A4B row input; CoT 42.5%
Claude Opus 4.6 row input; direct answer 40.6%
Gemma 4 26B A4B row input; direct answer 35.7%
GPT-5.4 row input; direct answer 33.5%
Mistral Small 3.1 row input; CoT 29.3%
Llama 3.3 70B row input; CoT 29.2%
Llama 3.3 70B row input; direct answer 27.2%
Gemma 3 27B row input; CoT 26.2%
Gemma 3 12B row input; CoT 24.6%
Mistral Small 3.1 row input; direct answer 24.5%
Llama 3.1-8B row input; CoT 23.8%
Gemma 3 27B row input; direct answer 23.8%
Llama 3.1-8B row input; direct answer 22.3%
Gemma 3 12B row input; direct answer 21.7%
Gemma 3 4B row input; direct answer 20.6%
Gemma 3 4B row input; CoT 19.8%
Llama 3.2 3B row input; CoT 19.6%
Llama 3.2 3B row input; direct answer 17.9%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3.1 Pro
row input; …
72.9%
Claude Opus 4.6
row input; … row input; …
60.2%
Claude Sonnet 4
row input; …
53.2%
GPT-5.4
row input; … row input; …
51.2%
Gemma 4 26B A4B
row input; … row input; …
42.5%
Mistral Small 3.1
row input; …
29.3%
Llama 3.3 70B
row input; …
29.2%
Gemma 3 27B
row input; …
26.2%
Gemma 3 12B
row input; …
24.6%
Llama 3.1-8B
row input; …
23.8%
Gemma 3 4B
row input; …
20.6%
Llama 3.2 3B
row input; …
19.6%

Source: The WearableQA paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro row input; CoT 72.9%
Gemini 3.1 Pro row input; direct answer 71.3%
Claude Opus 4.6 row input; CoT 60.2%
Claude Sonnet 4 row input; CoT 53.2%
GPT-5.4 row input; CoT 51.2%
Claude Sonnet 4 row input; direct answer 48.3%
Gemma 4 26B A4B row input; CoT 42.5%
Claude Opus 4.6 row input; direct answer 40.6%
Gemma 4 26B A4B row input; direct answer 35.7%
GPT-5.4 row input; direct answer 33.5%
Mistral Small 3.1 row input; CoT 29.3%
Llama 3.3 70B row input; CoT 29.2%
Llama 3.3 70B row input; direct answer 27.2%
Gemma 3 27B row input; CoT 26.2%
Gemma 3 12B row input; CoT 24.6%
Mistral Small 3.1 row input; direct answer 24.5%
Llama 3.1-8B row input; CoT 23.8%
Gemma 3 27B row input; direct answer 23.8%
Llama 3.1-8B row input; direct answer 22.3%
Gemma 3 12B row input; direct answer 21.7%
Gemma 3 4B row input; direct answer 20.6%
Gemma 3 4B row input; CoT 19.8%
Llama 3.2 3B row input; CoT 19.6%
Llama 3.2 3B row input; direct answer 17.9%

Number one over time

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

Google DeepMind
Benchmark paper Not tested since Sep 2026

Source: The WearableQA paper

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
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Pro Claude Opus 4.6 Claude Sonnet 4 Mistral Small 3.1 Llama 3.1-8B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Pro Claude Opus 4.6

Source: Official WearableQA 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (14 entries).
  • Pass Models scored 12 models we track, plus 2 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 (#28).
  • Pass Headroom The best is 72.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 24 rows for models we track and 2 other entries at the last read.
  • Pass Last read November 19th, 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