BrowseComp leaderboard
As of 6 Nov 2026, the top score on BrowseComp is 31.2% by Gemini 3.1 Pro (Google), ahead of GPT-5 at 20.1%, across 40 model configurations with a published result (OpenAI, result dated 25 Mar 2026). Hard questions whose answers are buried deep on the open web: the model has to browse, not recall. Published by OpenAI. BrowseComp is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- Gemini 3.1 Pro holds the top BrowseComp score at 31.2% (OpenAI, 25 Mar 2026).
- GPT-5 is second on BrowseComp at 20.1% (OpenAI, 26 Aug 2025).
- BrowseComp: Hard questions whose answers are buried deep on the open web: the model has to browse, not recall. Published by OpenAI (OpenAI).
- 40 model configurations have a published BrowseComp result on BenchLeader (BenchLeader, 6 Nov 2026).
- BrowseComp does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Hard questions whose answers are buried deep on the open web: the model has to browse, not recall. Published by OpenAI.
- Published by
- OpenAIdata via Kaggle Benchmarks
- Category
- Agents & tools
- Index weight
- Reference only
- Models
- 40
- Data as of
- 6 Nov 2026
Kaggle (kaggle.com); each board credits its publishing lab.
- 1Gemini 3.1 Pro31.2%
- 2GPT-520.1%
- 3Gemini 3 Flash16.9%
- 4Gemini 2.5 Pro7.8%
- 5o16.3%
- 6GPT-4.13.5%
- 7o4-mini3.3%
- 8DeepSeek R1 05283.1%
- 9Gemini 2.5 Flash2.5%
- 10Grok 3 mini1.7%
- 11gpt-oss-120b1.6%
- 12o3-mini1.5%
- 13Grok 31.3%
- 14DeepSeek V31.3%
- 15gpt-oss-20b1.3%
40 of 40
| # | ||||
|---|---|---|---|---|
| 1 |  Gemini 3.1 Pro | 31.2% | 60.5 | 2026-03-25 |
| 2 |  GPT-5 | 20.1% | 57.7 | 2025-08-26 |
| 3 | Gemini 3 Flash | 16.9% | 56.9 | 2026-03-23 |
| 4 | Gemini 2.5 Pro | 7.8% | 54.5 | 2025-06-13 |
| 5 | o1 | 6.3% | 53.4 | 2025-06-13 |
| 6 | GPT-4.1 | 3.5% | 48.5 | 2025-06-13 |
| 7 | o4-mini | 3.3% | 56.9 | 2025-06-13 |
| 8 |  DeepSeek R1 0528open ↗ | 3.1% | 49.3 | 2025-07-02 |
| 9 | Gemini 2.5 Flash | 2.5% | 52.7 | 2025-06-13 |
| 10 |  Grok 3 mini | 1.7% | 52.3 | 2025-06-13 |
| 11 | gpt-oss-120bopen ↗ | 1.6% | 45.5 | 2025-08-21 |
| 12 | o3-mini | 1.5% | 50.7 | 2025-06-13 |
| 13 | Grok 3 | 1.3% | 52.5 | 2025-06-13 |
| 14 | DeepSeek V3open ↗ | 1.3% | 44.4 | 2025-06-13 |
| 15 | gpt-oss-20bopen ↗ | 1.3% | 44.0 | 2025-08-21 |
| 16 | Gemini 3.1 Flash Lite | 1.1% | 52.4 | 2026-03-23 |
| 17 |  Claude Opus 4 | 1.0% | 52.6 | 2025-06-13 |
| 18 | Claude Opus 4.1 | 1.0% | 51.4 | 2025-09-03 |
| 19 | Grok 2open | 1.0% | 42.3 | 2025-06-13 |
| 20 | Claude 3.5 Sonnet | 0.9% | 47.1 | 2025-06-13 |
| 21 | o1-mini | 0.9% | 47.0 | 2025-06-13 |
| 22 | Claude 3.7 Sonnet | 0.9% | 49.9 | 2025-06-13 |
| 23 | GPT-4o | 0.7% | 43.3 | 2025-06-13 |
| 24 | Claude Sonnet 4 | 0.6% | 50.5 | 2025-06-13 |
| 25 | GPT 3.5 Turbo 1106 | 0.6% | 38.6 | 2025-06-13 |
| 26 | GPT-4o mini | 0.6% | 35.4 | 2025-06-13 |
| 27 | Gemini 2.0 Flash | 0.6% | 43.5 | 2025-06-13 |
| 28 |  Mistral Large 2.1open | 0.6% | 42.6 | 2025-07-02 |
| 29 | Gemma 3 27Bopen ↗ | 0.6% | 39.9 | 2025-06-13 |
| 30 |  Qwen3 235B A22B 2507thinkingopen ↗ | 0.5% | 51.9 | 2025-08-21 |
| 31 | Gemma 3 12Bopen ↗ | 0.5% | 37.9 | 2025-06-13 |
| 32 | Mixtral 8x22Bopen | 0.5% | – | 2025-07-02 |
| 33 |  Granite 4.0 H Smallopen ↗ | 0.4% | 42.6 | 2025-09-30 |
| 34 | Gemini 1.5 Pro 002 | 0.4% | 41.5 | 2025-06-13 |
| 35 | Gemini 1.5 Flash 002 | 0.4% | 38.2 | 2025-06-13 |
| 36 | Gemma 3 4Bopen ↗ | 0.4% | 34.9 | 2025-06-13 |
| 37 | Claude 3.5 Haiku | 0.3% | 38.8 | 2025-06-13 |
| 38 | Ministral 8Bopen | 0.3% | 37.8 | 2025-07-02 |
| 39 | Ministral 3Bopen | 0.2% | – | 2025-07-02 |
| 40 | Gemini 1.5 Flash 8B 001 | 0.1% | – | 2025-06-13 |
More agents & tools benchmarks · All LLM benchmarks
Related benchmarks
Cite as: BenchLeader, “BrowseComp leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/kaggle_browsecomp.
BrowseComp: questions
- What does BrowseComp measure?
- Hard questions whose answers are buried deep on the open web: the model has to browse, not recall. Published by OpenAI. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on BrowseComp?
- As of 6 Nov 2026, the highest score on BrowseComp is 31.2% by Gemini 3.1 Pro (result dated 25 Mar 2026), ahead of GPT-5 at 20.1%. Scores come from OpenAI and are re-read daily.
- Which AI model leads BrowseComp?
- Gemini 3.1 Pro leads BrowseComp with 31.2% as of 6 Nov 2026 (result dated 25 Mar 2026), ahead of GPT-5 at 20.1%.
- How many models have BrowseComp results?
- 40 model configurations have a BrowseComp result on BenchLeader, all taken from OpenAI via Kaggle Benchmarks.
- Who runs BrowseComp and how often is it updated?
- BrowseComp is published by OpenAI. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does BrowseComp count toward the BenchLeader Index?
- No. BrowseComp is shown for reference but left out of the composite index.