Parallel Search Capability Leaderboard (Parallel Search Capability)
We show this table for reference; we do not rank on it.
A 29-model comparison of how well models use Parallel Search Fast and Extract to answer web research questions.
Search Intelligence Score on Parallel Search Capability Leaderboard — Nov 10, 2026
We mirror the published search intelligence score view for Parallel Search Capability Leaderboard. Claude Fable 5.2 leads the public snapshot at 80.1, followed by Claude Opus 5.5 (76.1) and Hy4 (75.4). We do not use these results to rank models overall.
Claude Fable 5.2
Anthropic
Without search 56.9 · +23.2 lift · $2,426 / 1K tasks
Claude Opus 5.5
Anthropic
Without search 44.2 · +31.9 lift · $896 / 1K tasks · 214s / task
Hy4
Tencent
Without search 30.2 · +45.3 lift · $94.4 / 1K tasks · 250s / task
56 modelsAgenticCurrentDisplay onlyUpdated Nov 10, 2026
Search Intelligence Score table (56 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Fable 5.2AnthropicWithout search 56.9 · +23.2 lift · $2,426 / 1K tasks | 80.1 | Not reported | Closed |
| 2 | Claude Opus 5.5AnthropicWithout search 44.2 · +31.9 lift · $896 / 1K tasks · 214s / task | 76.1 | Not reported | Closed |
| 3 | Hy4TencentWithout search 30.2 · +45.3 lift · $94.4 / 1K tasks · 250s / task | 75.4 | Not reported | Pending |
| 4 | Gemini 4 CarbonGoogleWithout search 40.7 · +33.6 lift · $472 / 1K tasks · 1215s / task | 74.2 | Not reported | Closed |
| 5 | Kimi K3.1 ProMoonshot AIWithout search 38.8 · +35.5 lift · $268 / 1K tasks · 257s / task | 74.2 | Not reported | Open |
| 6 | Claude Sonnet 5.5AnthropicWithout search 33.2 · +40.9 lift · $796 / 1K tasks · 500s / task | 74.1 | Not reported | Closed |
| 7 | Claude Fable 5.1AnthropicWithout search 46.8 · +26.5 lift · $1,762 / 1K tasks · 416s / task | 73.4 | Not reported | Closed |
| 8 | Gemini 4 HydrogenGoogleWithout search 33.8 · +38.8 lift · $175 / 1K tasks | 72.7 | Not reported | Closed |
| 9 | Pareto 26.9UnbiasedWithout search 39.7 · +32.8 lift · $184 / 1K tasks · 211s / task | 72.5 | Not reported | Closed |
| 10 | Gemini 4 ArgonGoogleWithout search 34.9 · +37.2 lift · $270 / 1K tasks · 206s / task | 72.2 | Not reported | Closed |
| 11 | Muse Wonder 1.0UnknownWithout search 35.3 · +36.0 lift · $709 / 1K tasks · 287s / task | 71.3 | Not reported | Unknown |
| 12 | GPT-6 AstraOpenAIWithout search 43.0 · +27.9 lift · $401 / 1K tasks · 83.7s / task | 70.8 | Not reported | Closed |
| 13 | GPT-6.1 SolOpenAIWithout search 39.8 · +30.6 lift · $130 / 1K tasks · 353s / task | 70.4 | Not reported | Closed |
| 14 | Claude Opus 5AnthropicWithout search 37.2 · +32.8 lift · $1,012 / 1K tasks · 342s / task | 70.0 | Not reported | Closed |
| 15 | GPT-6.1 TerraOpenAIWithout search 28.6 · +41.2 lift · 42.3s / task | 69.7 | Not reported | Closed |
| 16 | OpenRouter Auto (max)Unknown$88.4 / 1K tasks · 97.5s / task | 69.2 | Not reported | Unknown |
| 17 | Particle-1Particle AIWithout search 31.4 · +37.8 lift · $48.4 / 1K tasks · 82.5s / task | 69.2 | Not reported | Closed |
| 18 | DeepSeek V4.1 Flash (1017)DeepSeekWithout search 32.9 · +36.0 lift · $125 / 1K tasks · 540s / task | 68.9 | Not reported | Unknown |
| 19 | MiniMax M3.1 FlashMiniMaxWithout search 27.1 · +41.7 lift · $43.8 / 1K tasks | 68.8 | Not reported | Closed |
| 20 | Pareto 26.10UnknownWithout search 34.8 · +33.8 lift · $116 / 1K tasks · 436s / task | 68.6 | Not reported | Unknown |
| 21 | Ember-1FireworksWithout search 29.9 · +38.0 lift · $335 / 1K tasks · 487s / task | 67.9 | Not reported | Closed |
| 22 | GPT-5.6 SolOpenAIWithout search 41.4 · +26.3 lift · $269 / 1K tasks · 107s / task | 67.7 | Not reported | Closed |
| 23 | Muse Spark OpenMetaWithout search 35.5 · +31.8 lift · $171 / 1K tasks · 360s / task | 67.2 | Not reported | Open |
| 24 | Gemini 3.7 FlashGoogleWithout search 39.4 · +27.4 lift · $130 / 1K tasks · 229s / task | 66.8 | Not reported | Closed |
| 25 | GPT-6 SolOpenAIWithout search 37.9 · +28.7 lift · $183 / 1K tasks · 260s / task | 66.6 | Not reported | Closed |
| 26 | Claude Sonnet 5AnthropicWithout search 22.8 · +43.8 lift · $688 / 1K tasks · 846s / task | 66.6 | Not reported | Closed |
| 27 | Muse Spark 1.3MetaWithout search 30.3 · +35.8 lift · $142 / 1K tasks · 274s / task | 66.1 | Not reported | Closed |
| 28 | Qwen3.8 MaxAlibabaWithout search 28.9 · +37.1 lift · $221 / 1K tasks | 66.0 | Not reported | Open |
| 29 | Grok 4.7 MediumxAIWithout search 21.1 · +44.8 lift · $252 / 1K tasks · 521s / task | 66.0 | Not reported | Unknown |
| 30 | Gemini 3.8 FlashGoogleWithout search 39.0 · +26.1 lift · $177 / 1K tasks · 317s / task | 65.1 | Not reported | Closed |
| 31 | Hy4 PreviewUnknownWithout search 30.6 · +33.9 lift · $93.2 / 1K tasks · 722s / task | 64.6 | Not reported | Unknown |
| 32 | Mistral Large 4.0MistralWithout search 21.4 · +43.0 lift · $136 / 1K tasks | 64.4 | Not reported | Unknown |
| 33 | MiMo v2.6 ProUnknownWithout search 29.7 · +34.7 lift · $76.5 / 1K tasks | 64.4 | Not reported | Unknown |
| 34 | Kimi K3Moonshot AIWithout search 30.6 · +33.6 lift · $266 / 1K tasks · 432s / task | 64.2 | Not reported | Open |
| 35 | Mercury 3InceptionWithout search 28.1 · +35.5 lift · $43.4 / 1K tasks · 149s / task | 63.6 | Not reported | Closed |
| 36 | Claude Haiku 5.5AnthropicWithout search 23.5 · +39.4 lift · $24.3 / 1K tasks · 162s / task | 62.9 | Not reported | Closed |
| 37 | DeepSeek V4.1 FlashDeepSeekWithout search 27.0 · +35.7 lift · $35.8 / 1K tasks · 549s / task | 62.7 | Not reported | Unknown |
| 38 | GLM 5.3Z.AIWithout search 21.9 · +40.8 lift · $78.5 / 1K tasks · 308s / task | 62.7 | Not reported | Unknown |
| 39 | GPT-6 LunaOpenAIWithout search 28.0 · +33.9 lift · $33.1 / 1K tasks · 391s / task | 61.9 | Not reported | Closed |
| 40 | Step 5 PreviewStepFunWithout search 27.8 · +33.3 lift · $118 / 1K tasks · 387s / task | 61.1 | Not reported | Open |
| 41 | Ultramarine 1 FlashPaper InstrumentsWithout search 24.0 · +37.1 lift · $87.2 / 1K tasks · 165s / task | 61.1 | Not reported | Closed |
| 42 | MiMo v2.6 FlashUnknownWithout search 22.4 · +38.5 lift · $25.2 / 1K tasks · 1323s / task | 60.9 | Not reported | Unknown |
| 43 | GLM 5.3 FlashZ.AIWithout search 20.4 · +40.4 lift · $16.6 / 1K tasks · 99.1s / task | 60.8 | Not reported | Unknown |
| 44 | GPT-5.6 LunaOpenAIWithout search 26.9 · +33.9 lift · $36.2 / 1K tasks · 98s / task | 60.7 | Not reported | Closed |
| 45 | Nemotron 3.5 Super VLNVIDIAWithout search 17.0 · +42.2 lift · $93.9 / 1K tasks · 181s / task | 59.1 | Not reported | Unknown |
| 46 | Gemini 3 FlashGoogleWithout search 34.1 · +24.5 lift · $114 / 1K tasks · 276s / task | 58.6 | Not reported | Closed |
| 47 | DeepSeek V4 ProDeepSeekWithout search 30.5 · +27.9 lift · $110 / 1K tasks · 383s / task | 58.4 | Not reported | Unknown |
| 48 | Hunyuan 3UnknownWithout search 22.6 · +35.6 lift · $28.4 / 1K tasks · 363s / task | 58.2 | Not reported | Unknown |
| 49 | BeamUnknownWithout search 28.8 · +27.2 lift · $161 / 1K tasks | 56.0 | Not reported | Unknown |
| 50 | GLM-5.2Z.AIWithout search 18.4 · +35.9 lift · $66.6 / 1K tasks · 301s / task | 54.3 | Not reported | Open |
| 51 | Nemotron 3 Ultra 550BNVIDIAWithout search 15.3 · +37.7 lift · $113 / 1K tasks · 328s / task | 53.0 | Not reported | Unknown |
| 52 | MiniMax M3MiniMaxWithout search 23.1 · +29.5 lift · $41.1 / 1K tasks · 291s / task | 52.6 | Not reported | Open |
| 53 | MiMo v2.5UnknownWithout search 13.4 · +36.1 lift · $23.5 / 1K tasks · 894s / task | 49.5 | Not reported | Unknown |
| 54 | Laguna S 2.1PoolsideWithout search 12.3 · +27.2 lift · $21.4 / 1K tasks · 459s / task | 39.5 | Not reported | Open |
| 55 | Nemotron 3.5 LightningNVIDIAWithout search 9.8 · +28.0 lift · $12.1 / 1K tasks · 176s / task | 37.9 | Not reported | Unknown |
| 56 | Command A PlusCohereWithout search 5.0 · +29.8 lift · $40.1 / 1K tasks · 149s / task | 34.8 | Not reported | Unknown |
How to read this leaderboard
Compare the models within Parallel's fixed tool setup. Search lift shows the difference between the search and no-search runs; cost is reported per 1,000 tasks and includes inference plus estimated Search and Extract usage.
Operator receipt: 56 sourced rows are currently displayable on this page; the leading published row is Claude Fable 5.2 at 80.1.
Honest limit: The composite includes Parallel’s proprietary WISER set and uses Parallel’s search tools for every model. Reasoning settings and token limits can vary by model. Failed and pending tasks count as zero across all 100 questions per component. Cost records are incomplete and may omit recovery attempts. The table remains display only and does not enter overall or category rankings.
How to read the search capability table
We captured 56 rows from Parallel's Nov 10, 2026 Search Capability Leaderboard. The score equally weights DeepSearchQA F1, Humanity's Last Exam accuracy, and WISER accuracy. Each model was tested on 100 questions from each set with Parallel Search Fast and Extract.
The score reflects a model using Parallel's search tools in this test setup. The row labels show the no-search score, lift, cost per 1,000 tasks, and time per task. Failed and pending tasks count as zero; cost records may omit recovery attempts. The table is display only and does not enter overall or category rankings.
Snapshot
The published Parallel Search Capability snapshot places Claude Fable 5.2 first at 80.1. The third row is 4.7 score units behind. The broader top-10 range is 7.9 score units, so many of the published results sit in a relatively narrow band.
56 models have been evaluated on Parallel Search Capability. The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Parallel Search Capability is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Parallel Search Capability
Year
2026
Tasks
100 questions each from DeepSearchQA, HLE, and WISER
Format
Equal-weight Search Intelligence Score (0–100)
Difficulty
Multi-step web research and specialist questions
Parallel runs 100 questions from each of DeepSearchQA, Humanity's Last Exam, and WISER with and without Parallel Search Fast and Extract. The Search Intelligence Score equally weights DSQA F1, HLE accuracy, and WISER accuracy. We mirror the 29 published composite scores and preserve no-search scores, search lift, cost, and time in the source snapshot.
Freshness and provenance
Version
Parallel Search Capability 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
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 Parallel Search Capability measure?
A 29-model comparison of how well models use Parallel Search Fast and Extract to answer web research questions.
Which model leads the published Parallel Search Capability snapshot?
Claude Fable 5.2 currently leads the published Parallel Search Capability snapshot with 80.1 search intelligence score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on Parallel Search Capability?
The Nov 10, 2026 snapshot contains 56 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.