Agents · Benchmark profile
An agentic browsing benchmark where models search the web, gather evidence, and answer list-style questions using browser tools.
Data verified 21 Jul 2026 · Methodology 1.6.0
Benchmark score on DeepSearchQA
Kimi K3 leads at 95%, followed by Claude Opus 4.8 (93.1%) and Step 3.7 Flash (92.8%).
Visual analysis
Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.
11 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Kimi K3 | #1 | Moonshot AI | 95% |
| Claude Opus 4.8 | #2 | Anthropic | 93.1% |
| Step 3.7 Flash | #3 | StepFun | 92.8% |
| Kimi K2.6 | #4 | Moonshot AI | 92.5% |
| Muse Spark 1.1 | #5 | Meta | 84.9% |
| Kimi K2.5 | #6 | Moonshot AI | 77.1% |
| Muse Spark | #7 | Meta | 74.8% |
| Claude Opus 4.6 | #8 | Anthropic | 73.7% |
| GPT-5.4 | #9 | OpenAI | 73.6% |
| Gemini 3.1 Pro | #10 | 69.7% | |
| Grok 4.20 | #11 | xAI | 62.8% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | Kimi K3 kimi-k3 | Moonshot AI | closed | Reference only | 95% |
| #2 | Claude Opus 4.8 claude-opus-4-8 | Anthropic | closed | Reference only | 93.1% |
| #3 | Step 3.7 Flash step-3-7-flash | StepFun | open | Estimated reference | 92.8% |
| #4 | Kimi K2.6 kimi-k2.6 | Moonshot AI | open | Estimated reference | 92.5% |
| #5 | Muse Spark 1.1 muse-spark-1-1 | Meta | closed | Reference only | 84.9% |
| #6 | Kimi K2.5 kimi-k2.5 | Moonshot AI | — | Reference only | 77.1% |
| #7 | Muse Spark muse-spark | Meta | closed | Reference only | 74.8% |
| #8 | Claude Opus 4.6 claude-opus-4-6 | Anthropic | closed | Reference only | 73.7% |
| #9 | GPT-5.4 gpt-5.4 | OpenAI | closed | Reference only | 73.6% |
| #10 | Gemini 3.1 Pro gemini-3-1-pro | closed | Estimated reference | 69.7% | |
| #11 | Grok 4.20 grok-4.20 | xAI | closed | Estimated reference | 62.8% |
The top of this snapshot is led by Kimi K3 at 95%; third place is 2.2 points behind. The top-10 spread is 25.3 points.
About DeepSearchQA
An agentic browsing benchmark where models search the web, gather evidence, and answer list-style questions using browser tools. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
An agentic browsing benchmark where models search the web, gather evidence, and answer list-style questions using browser tools.
Kimi K3 by Moonshot AI currently leads with 95%.
11 models in the LuminaBench cohort have a qualifying score on this benchmark.
No. This benchmark is display-only and does not enter the overall Lumina composite.
Related
