DeepSearchQA leaderboard
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Atria Dawn PreviewShanghai Artificial Intelligence Laboratory	96.0%	Not reported	Open
2	Claude Opus 5Anthropic	95.0%	Not reported	Closed
3	Kimi K3Moonshot AI	95.0%	Not reported	Open
4	Claude Opus 4.8Anthropic	93.1%	Not reported	Closed
5	Step 3.7 FlashStepFun	92.8%	Not reported	Open
6	Kimi K2.6Moonshot AI	92.5%	Not reported	Open
7	Apodex 1.1Apodex	92.4%	Not reported	Closed
8	dots3-note PreviewDots Studio	92.1%	Not reported	Open
9	Inkling 2 SmallThinking Machines Lab	90.9%	Not reported	Closed
10	Ling 3.0 Flash SanteInclusionAI	90.8%	Not reported	Open
11	Muse Spark 1.3Meta	89.4%	Not reported	Closed
12	Muse Spark 1.1Meta	84.9%	Not reported	Closed
13	Beam 1.0 501B-A23B NVFP4Reflection AI	81.1%	Not reported	Open
14	Beam 1.0 501B-A23BReflection AI	80.5%	Not reported	Open
15	Beam 1.0 501B-A23B FP8Reflection AI	79.6%	Not reported	Open
16	Kimi K2.5Moonshot AI	77.1%	Not reported	Open
17	Muse SparkMeta	74.8%	Not reported	Closed
18	Muse Glimmer 30BMeta	74.6%	Not reported	Open
19	Claude Opus 4.6Anthropic	73.7%	Not reported	Closed
20	GPT-5.4OpenAI	73.6%	Not reported	Closed
21	Mercury 3Inception	71.3%	Not reported	Closed
22	Gemini 3.1 ProGoogle	69.7%	Not reported	Closed
23	Grok 4.20xAI	62.8%	Not reported	Closed
24	Mercury 2.5Inception	34.0%	Not reported	Closed
25	Mercury 2Inception	34.0%	Not reported	Closed

According to BenchLM.ai, Atria Dawn Preview leads the DeepSearchQA benchmark with a score of 96.0%, followed by Claude Opus 5 (95.0%) and Kimi K3 (95.0%). The top models are clustered within 1.0 points, suggesting this benchmark is nearing saturation for frontier models.

25 models have been evaluated on DeepSearchQA. The benchmark falls in the Agentic category. BenchAlign v5.8 gives DeepSearchQA 2% of the Agentic reference weight, so it moves the Agentic leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.

About DeepSearchQA

Year

2026

Tasks

Agentic browsing and list-answer questions

Format

Search / open / find browser-agent evaluation

Difficulty

Agentic web research

Meta describes DeepSearchQA as a browser-tool evaluation graded with an F1-style semantic set match. BenchLM stores it as an agentic search benchmark.

Muse Spark Eval Methodology

Freshness and provenance

Version

DeepSearchQA 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does DeepSearchQA measure?

An agentic browsing benchmark where models search the web, gather evidence, and answer list-style questions using browser tools.

Which model scores highest on DeepSearchQA?

Atria Dawn Preview by Shanghai Artificial Intelligence Laboratory currently leads with a score of 96.0% on DeepSearchQA.

How many models are evaluated on DeepSearchQA?

25 AI models have published results on DeepSearchQA in the BenchLM catalog.

Compare top models on DeepSearchQA

Atria Dawn Preview vs Claude Opus 5 Claude Opus 5 vs Kimi K3 Kimi K3 vs Claude Opus 4.8 Claude Opus 4.8 vs Step 3.7 Flash

Last updated: November 11, 2026 · BenchLM version DeepSearchQA 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.