AI Model Release Tracker
Web browsing

BrowseComp

Can the AI browse the web and track down hard-to-find answers? Higher is better.

Rankings

Higher is better
#1
Kimi K3
91.2%
2
Claude Opus 5
90.8%
3
GPT-5.5-Pro
90.1%
4
GPT-5.4-Pro
89.3%
5
Claude Fable 5
86.9%
6
Gemini 3.1 Pro
85.9%
7
Claude Sonnet 5
84.7%
8
GPT-5.5
84.4%
9
Claude Opus 4.8
84.3%
10
Claude Opus 4.6
83.7%
11
DeepSeek-V4-Pro
83.4%
12
GPT-5.4
82.7%
13
Claude Opus 4.7
79.3%
14
GLM-5
75.9%
15
Qwen3.5OW
69%
16
GLM-5.1
68%
17
GLM-4.7
52%

BrowseComp — frequently asked questions

What is BrowseComp?
Can the AI browse the web and track down hard-to-find answers? Higher is better.
Which AI model scores highest on BrowseComp?
Kimi K3 by Moonshot AI holds the best BrowseComp result among tracked models, at 91.2% (released Jul 16 2026). Higher scores are better on this benchmark.
What are the top 5 models on BrowseComp?
1. Kimi K3 (Moonshot AI) — 91.2%; 2. Claude Opus 5 (Anthropic) — 90.8%; 3. GPT-5.5-Pro (OpenAI) — 90.1%; 4. GPT-5.4-Pro (OpenAI) — 89.3%; 5. Claude Fable 5 (Anthropic) — 86.9%.
How many models have a published BrowseComp score?
17 tracked models have a published BrowseComp score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.
What is the best open model on BrowseComp?
Qwen3.5 by Qwen is the highest-ranked model with downloadable weights on BrowseComp, scoring 69% at rank 15 overall.
← All benchmarks