math benchmark
DROP Leaderboard
DROP (Discrete Reasoning Over Paragraphs) is a reading comprehension benchmark with crowdsourced, adversarially-created questions requiring discrete reasoning over paragraph content.
Updated Sep 5, 2026
Models30
Model coverage30
MetricScore
EvidenceB
DROP Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelDE | Score91.60% | Percentile100.00% | Participants30 | EvidenceC | Evaluated |
| Rank02 | ModelAN | Score87.10% | Percentile96.55% | Participants30 | EvidenceC | Evaluated |
| Rank03 | ModelAN | Score87.10% | Percentile93.10% | Participants30 | EvidenceC | Evaluated |
| Rank04 | ModelXI | Score86.30% | Percentile89.66% | Participants30 | EvidenceC | Evaluated |
| Rank05 | ModelOP | Score86.00% | Percentile86.21% | Participants30 | EvidenceC | Evaluated |
| Rank06 | ModelAM Nova ProAmazon | Score85.40% | Percentile82.76% | Participants30 | EvidenceC | Evaluated |
| Rank07 | ModelME | Score84.80% | Percentile79.31% | Participants30 | EvidenceC | Evaluated |
| Rank08 | ModelOP | Score83.40% | Percentile75.86% | Participants30 | EvidenceC | Evaluated |
| Rank09 | ModelAN | Score83.10% | Percentile72.41% | Participants30 | EvidenceC | Evaluated |
| Rank10 | ModelAN | Score83.10% | Percentile68.97% | Participants30 | EvidenceC | Evaluated |
| Rank11 | ModelOP | Score80.90% | Percentile65.52% | Participants30 | EvidenceC | Evaluated |
| Rank12 | ModelAMNova LiteAmazon | Score80.20% | Percentile62.07% | Participants30 | EvidenceC | Evaluated |
| Rank13 | ModelOP | Score79.70% | Percentile58.62% | Participants30 | EvidenceC | Evaluated |
| Rank14 | ModelME | Score79.60% | Percentile55.17% | Participants30 | EvidenceC | Evaluated |
| Rank15 | ModelAMNova MicroAmazon | Score79.30% | Percentile51.72% | Participants30 | EvidenceC | Evaluated |
| Rank16 | ModelME | Score79.06% | Percentile48.28% | Participants30 | EvidenceC | Evaluated |
| Rank17 | ModelAN | Score78.90% | Percentile44.83% | Participants30 | EvidenceC | Evaluated |
| Rank18 | ModelAN | Score78.40% | Percentile41.38% | Participants30 | EvidenceC | Evaluated |
| Rank19 | ModelMI | Score75.50% | Percentile37.93% | Participants30 | EvidenceC | Evaluated |
| Rank20 | ModelGO | Score74.90% | Percentile34.48% | Participants30 | EvidenceC | Evaluated |
| Rank21 | ModelOP | Score70.20% | Percentile31.03% | Participants30 | EvidenceB | Evaluated |
| Rank22 | ModelGO | Score60.80% | Percentile27.59% | Participants30 | EvidenceC | Evaluated |
| Rank23 | ModelGO | Score60.80% | Percentile24.14% | Participants30 | EvidenceC | Evaluated |
| Rank24 | ModelME | Score59.50% | Percentile20.69% | Participants30 | EvidenceC | Evaluated |
| Rank25 | ModelIB | Score59.36% | Percentile17.24% | Participants30 | EvidenceC | Evaluated |
| Rank26 | ModelGO | Score53.90% | Percentile13.79% | Participants30 | EvidenceC | Evaluated |
| Rank27 | ModelGO | Score53.90% | Percentile10.34% | Participants30 | EvidenceC | Evaluated |
| Rank28 | ModelIB | Score46.20% | Percentile6.90% | Participants30 | EvidenceC | Evaluated |
| Rank29 | ModelIB | Score36.14% | Percentile3.45% | Participants30 | EvidenceC | Evaluated |
| Rank30 | ModelBA | Score28.60% | Percentile0.00% | Participants30 | EvidenceC | Evaluated |
DROP Highlights
The leading models and scores on this benchmark.
DROP Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for DROP
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is DROP?
What DROP measures and how its scores work.
DROP is a reading comprehension benchmark focused on discrete reasoning over paragraphs.
It measures performance on questions requiring reference resolution and discrete operations such as addition, counting, or sorting, along with comprehensive paragraph understanding.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- DROP
- Modality
- text
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about DROP.
Which model scores highest on DROP?
DeepSeek-V3 is currently ranked first with 91.60%.
What are the top three models on DROP?
The current leaders are DeepSeek-V3 (91.60%), Claude 3.5 Sonnet (87.10%), and Claude 3.5 Sonnet (87.10%).
Which DROP model has the lowest official input price?
Nova Micro has the lowest matched official input price at $0.04 input / $0.14 output per 1M tokens.
Which models are fastest among DROP results?
The fastest matched records are Llama 3.1 8B Instruct (2,047.00 tok/s via Cerebras), Llama 3.1 70B Instruct (1,204.00 tok/s via Cerebras), and GPT-4 (104.00 tok/s via Azure).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.