reasoning benchmark
BrowseComp Leaderboard
BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information. The benchmark measures agents' ability to exercise persistence in information gathering, demonstrate creativity in web navigation, and find concise, verifiable answers. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers.
Updated Aug 17, 2026
Models62
Model coverage62
MetricScore
EvidenceB
BrowseComp Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelMA | Score91.2% | Percentile100.0% | Participants62 | EvidenceC | Evaluated |
| Rank02 | ModelAN | Score90.8% | Percentile98.4% | Participants62 | EvidenceC | Evaluated |
| Rank03 | ModelOP | Score90.4% | Percentile96.7% | Participants62 | EvidenceC | Evaluated |
| Rank04 | ModelOP | Score90.1% | Percentile95.1% | Participants62 | EvidenceC | Evaluated |
| Rank05 | ModelOP | Score87.5% | Percentile93.4% | Participants62 | EvidenceC | Evaluated |
| Rank06 | ModelAN | Score86.9% | Percentile91.8% | Participants62 | EvidenceC | Evaluated |
| Rank07 | ModelMA | Score86.3% | Percentile90.2% | Participants62 | EvidenceC | Evaluated |
| Rank08 | ModelBY | Score86.2% | Percentile88.5% | Participants62 | EvidenceC | Evaluated |
| Rank09 | ModelGO | Score85.9% | Percentile86.9% | Participants62 | EvidenceC | Evaluated |
| Rank10 | ModelBY | Score84.9% | Percentile85.3% | Participants62 | EvidenceC | Evaluated |
| Rank11 | ModelAN | Score84.7% | Percentile83.6% | Participants62 | EvidenceC | Evaluated |
| Rank12 | ModelOP | Score84.4% | Percentile82.0% | Participants62 | EvidenceC | Evaluated |
| Rank13 | ModelAN | Score84.3% | Percentile80.3% | Participants62 | EvidenceC | Evaluated |
| Rank14 | ModelTE | Score84.2% | Percentile78.7% | Participants62 | EvidenceC | Evaluated |
| Rank15 | ModelAN | Score84.0% | Percentile77.0% | Participants62 | EvidenceC | Evaluated |
| Rank16 | ModelMI | Score83.5% | Percentile75.4% | Participants62 | EvidenceC | Evaluated |
| Rank17 | ModelDE | Score83.4% | Percentile73.8% | Participants62 | EvidenceC | Evaluated |
| Rank18 | ModelOP | Score83.3% | Percentile72.1% | Participants62 | EvidenceC | Evaluated |
| Rank19 | ModelOP | Score82.7% | Percentile70.5% | Participants62 | EvidenceC | Evaluated |
| Rank20 | ModelAN | Score79.3% | Percentile68.8% | Participants62 | EvidenceC | Evaluated |
| Rank21 | ModelZA | Score79.3% | Percentile67.2% | Participants62 | EvidenceC | Evaluated |
| Rank22 | ModelOP | Score77.9% | Percentile65.6% | Participants62 | EvidenceC | Evaluated |
| Rank23 | ModelTM | Score77.4% | Percentile63.9% | Participants62 | EvidenceC | Evaluated |
| Rank24 | ModelBY | Score77.3% | Percentile62.3% | Participants62 | EvidenceC | Evaluated |
| Rank25 | ModelMI | Score76.3% | Percentile60.7% | Participants62 | EvidenceC | Evaluated |
| Rank26 | ModelZA | Score75.9% | Percentile59.0% | Participants62 | EvidenceC | Evaluated |
| Rank27 | ModelMA | Score74.9% | Percentile57.4% | Participants62 | EvidenceC | Evaluated |
| Rank28 | ModelAN | Score74.7% | Percentile55.7% | Participants62 | EvidenceC | Evaluated |
| Rank29 | ModelDE | Score73.2% | Percentile54.1% | Participants62 | EvidenceC | Evaluated |
| Rank30 | ModelAC | Score69.0% | Percentile52.5% | Participants62 | EvidenceC | Evaluated |
| Rank31 | ModelST | Score69.0% | Percentile50.8% | Participants62 | EvidenceC | Evaluated |
| Rank32 | ModelOP | Score65.8% | Percentile49.2% | Participants62 | EvidenceC | Evaluated |
| Rank33 | ModelAC | Score63.8% | Percentile47.5% | Participants62 | EvidenceC | Evaluated |
| Rank34 | ModelMI | Score62.0% | Percentile45.9% | Participants62 | EvidenceC | Evaluated |
| Rank35 | ModelAC | Score61.0% | Percentile44.3% | Participants62 | EvidenceC | Evaluated |
| Rank36 | ModelAC | Score61.0% | Percentile42.6% | Participants62 | EvidenceC | Evaluated |
| Rank37 | ModelMA | Score60.2% | Percentile41.0% | Participants62 | EvidenceC | Evaluated |
| Rank38 | ModelXI | Score58.3% | Percentile39.3% | Participants62 | EvidenceC | Evaluated |
| Rank39 | ModelME | Score56.6% | Percentile37.7% | Participants62 | EvidenceC | Evaluated |
| Rank40 | ModelOP | Score54.9% | Percentile36.1% | Participants62 | EvidenceC | Evaluated |
| Rank41 | ModelDE | Score53.5% | Percentile34.4% | Participants62 | EvidenceC | Evaluated |
| Rank42 | ModelZA | Score52.0% | Percentile32.8% | Participants62 | EvidenceC | Evaluated |
| Rank43 | ModelOP | Score51.5% | Percentile31.1% | Participants62 | EvidenceC | Evaluated |
| Rank44 | ModelDE | Score51.4% | Percentile29.5% | Participants62 | EvidenceC | Evaluated |
| Rank45 | ModelDE | Score51.4% | Percentile27.9% | Participants62 | EvidenceC | Evaluated |
| Rank46 | ModelOP | Score49.7% | Percentile26.2% | Participants62 | EvidenceC | Evaluated |
| Rank47 | ModelSA | Score49.5% | Percentile24.6% | Participants62 | EvidenceC | Evaluated |
| Rank48 | ModelUP | Score49.2% | Percentile22.9% | Participants62 | EvidenceC | Evaluated |
| Rank49 | ModelMA | Score48.6% | Percentile21.3% | Participants62 | EvidenceC | Evaluated |
| Rank50 | ModelZA | Score45.1% | Percentile19.7% | Participants62 | EvidenceC | Evaluated |
| Rank51 | ModelXA | Score44.9% | Percentile18.0% | Participants62 | EvidenceC | Evaluated |
| Rank52 | ModelNV | Score44.4% | Percentile16.4% | Participants62 | EvidenceC | Evaluated |
| Rank53 | ModelMI | Score44.0% | Percentile14.8% | Participants62 | EvidenceC | Evaluated |
| Rank54 | ModelZA | Score42.8% | Percentile13.1% | Participants62 | EvidenceC | Evaluated |
| Rank55 | ModelDE | Score40.1% | Percentile11.5% | Participants62 | EvidenceC | Evaluated |
| Rank56 | ModelNV | Score37.0% | Percentile9.8% | Participants62 | EvidenceC | Evaluated |
| Rank57 | ModelSA | Score35.5% | Percentile8.2% | Participants62 | EvidenceC | Evaluated |
| Rank58 | ModelNV | Score31.3% | Percentile6.6% | Participants62 | EvidenceC | Evaluated |
| Rank59 | ModelDE | Score30.0% | Percentile4.9% | Participants62 | EvidenceC | Evaluated |
| Rank60 | ModelZA | Score26.4% | Percentile3.3% | Participants62 | EvidenceC | Evaluated |
| Rank61 | ModelZA | Score21.3% | Percentile1.6% | Participants62 | EvidenceC | Evaluated |
| Rank62 | ModelDE | Score8.9% | Percentile0.0% | Participants62 | EvidenceC | Evaluated |
BrowseComp Highlights
The leading models and scores on this benchmark.
BrowseComp Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for BrowseComp
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis browsecomp AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01MAKimi K3Moonshot AI
- Score
- 91.2%
- Price
- $3.0 input / $15 output per 1M tokens
- Speed
- Up to 26 tok/s via Fireworks
Strengths
- Ranks #1 of 62 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp, not total model capability
- 02ANClaude Opus 5Anthropic
- Score
- 90.8%
- Price
- $5.0 input / $25 output per 1M tokens
- Speed
- Up to 3.3 tok/s via Anthropic
Strengths
- Ranks #2 of 62 compared models
- 98th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp, not total model capability
- 03OPGPT-5.6 SolOpenAI
- Score
- 90.4%
- Price
- $5.0 input / $30 output per 1M tokens
- Speed
- Up to 27 tok/s via OpenAI
Strengths
- Ranks #3 of 62 compared models
- 97th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp, not total model capability
- 04OPGPT-5.5 ProOpenAI
- Score
- 90.1%
- Price
- $30 input / $180 output per 1M tokens
Strengths
- Ranks #4 of 62 compared models
- 95th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp, not total model capability
- 05OPGPT-5.6 TerraOpenAI
- Score
- 87.5%
- Price
- $2.0 input / $12 output per 1M tokens
- Speed
- Up to 35 tok/s via OpenAI
Strengths
- Ranks #5 of 62 compared models
- 93th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp, not total model capability
Selection summary
Best AI Models for BrowseComp
Kimi K3 currently leads BrowseComp with 91.2%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is BrowseComp?
What BrowseComp measures and how its scores work.
BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information. The benchmark measures agents' ability to exercise persistence in information gathering, demonstrate creativity in web navigation, and find concise, verifiable answers. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- BrowseComp
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- browsecomp|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about BrowseComp.
Which model scores highest on BrowseComp?
Kimi K3 is currently ranked first with 91.2%.
What does BrowseComp measure?
BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information. The benchmark measures agents' ability to exercise persistence in information gathering, demonstrate creativity in web navigation, and find concise, verifiable answers. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
62 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.