FACTS Search V1
We show this table for reference; we do not rank on it.
The archived FACTS Search V1 protocol measures factual answering with Brave Search and a Gemini 2.0 Flash grader. Its scores remain separate from Search V2.
Evaluation results published by Google DeepMind, Google Research, and Kaggle. How Kaggle Benchmarks reruns tests.
The exact version 1 page declares Apache 2.0. We retain aggregate measurements, attribution, and source links. No tasks, questions, answers, documents, or other underlying dataset contents are included. Original article terms.
Published score (%) on FACTS Search V1 — Archived version 1
We mirror the published published score (%) view for FACTS Search V1. Gemini 3.1 Pro Preview has the highest reported value at 85.6%. These archived version 1 protocols remain separate from the current version 2 boards. The complete official download includes configurations hidden by default source filters. Source identities are preserved, and these measurements do not enter overall or category rankings. We do not use these results to rank models overall.
Gemini 3.1 Pro Preview
Archived Kaggle benchmark v1; Task revision 1; Evaluation date unreported
GPT-5.5
Archived Kaggle benchmark v1; Task revision 1; Evaluation date unreported
Gemini 3 Flash Preview
Archived Kaggle benchmark v1; Task revision 1; Evaluation date unreported
31 source rowsAgenticCurrentDisplay onlyUpdated Archived version 1
Published score (%) table (31 source rows)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Gemini 3.1 Pro PreviewCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 85.6% | Not reported | Unknown |
| 2 | GPT-5.5Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 85.0% | Not reported | Unknown |
| 3 | Gemini 3 Flash PreviewCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 81.0% | Not reported | Unknown |
| 4 | Claude Opus 4.8Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 80.3% | Not reported | Unknown |
| 5 | Claude Opus 4.7Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 78.8% | Not reported | Unknown |
| 6 | GPT-5Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 77.7% | Not reported | Unknown |
| 7 | Grok 4.1 Fast ReasoningCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 77.4% | Not reported | Unknown |
| 8 | Claude Opus 4.6Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 76.7% | Not reported | Unknown |
| 9 | Claude Sonnet 4.6Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 76.4% | Not reported | Unknown |
| 10 | Grok 4Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 75.3% | Not reported | Unknown |
| 11 | o3Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 74.8% | Not reported | Unknown |
| 12 | Claude Opus 4.5Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 73.2% | Not reported | Unknown |
| 13 | GPT-5.2Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 72.2% | Not reported | Unknown |
| 14 | Claude Sonnet 4.5 (thinking)Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 69.8% | Not reported | Unknown |
| 15 | Gemma 4 31BCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 68.9% | Not reported | Unknown |
| 16 | GPT-5 miniCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 67.9% | Not reported | Unknown |
| 17 | grok-4-fast-reasoningCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 67.3% | Not reported | Unknown |
| 18 | Gemini 3.1 Flash-Lite PreviewCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 66.8% | Not reported | Unknown |
| 19 | Claude Sonnet 4Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 66.3% | Not reported | Unknown |
| 20 | o4 miniCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 66.2% | Not reported | Unknown |
| 21 | Claude Opus 4.1Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 65.0% | Not reported | Unknown |
| 22 | GPT-4.1Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 64.6% | Not reported | Unknown |
| 23 | Gemini 2.5 ProCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 63.9% | Not reported | Unknown |
| 24 | Gemini 3.5 FlashCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 62.6% | Not reported | Unknown |
| 25 | GPT-5.1Creator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 62.4% | Not reported | Unknown |
| 26 | Gemini 2.5 FlashCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 60.0% | Not reported | Unknown |
| 27 | GPT-4oCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 56.8% | Not reported | Unknown |
| 28 | Gemma 4 26B A4BCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 56.7% | Not reported | Unknown |
| 29 | gpt-oss-120bCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-24 | 44.5% | Not reported | Unknown |
| 30 | Gemma 3 27BCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluation date unreported | 41.4% | Not reported | Unknown |
| 31 | gpt-oss-20bCreator not reportedArchived Kaggle benchmark v1; Task revision 1; Evaluated 2025-11-25 | 34.8% | Not reported | Unknown |
How to read FACTS Search V1
FACTS Search V1 gives the model a shared Brave Search tool for up to seven agent turns. A Gemini 2.0 Flash grader compares the final answer with the question and reference answer. Version 2 changes the grader, so its results belong to a separate board.
The download supplies Score, public and private scores, hedging rate, average searches, attempted accuracy, and F1. We display the published Score and retain the other measurements separately. The source describes an F1 measure, but its distinct Score and F1 fields are not treated as interchangeable.
The download was acquired on the recorded retrieval date. Per-result evaluation timestamps are publisher dates, not evidence of when this leaderboard snapshot was first available.
These archived version 1 protocols remain separate from the current version 2 boards. The complete official download includes configurations hidden by default source filters. Source identities are preserved, and these measurements do not enter overall or category rankings.
Snapshot
These archived version 1 protocols remain separate from the current version 2 boards. The complete official download includes configurations hidden by default source filters. Source identities are preserved, and these measurements do not enter overall or category rankings. The table retains each source configuration and its reported value.
31 source rows are shown for FACTS Search V1. 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. FACTS Search V1 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About FACTS Search V1
Year
2025
Tasks
Search-assisted factual questions; public and held-out sets
Format
Tool-assisted answering with an automated grader
FACTS Search V1 gives the model a shared Brave Search tool for up to seven agent turns. A Gemini 2.0 Flash grader compares the final answer with the question and reference answer. Version 2 changes the grader, so its results belong to a separate board. The download supplies Score, public and private scores, hedging rate, average searches, attempted accuracy, and F1. We display the published Score and retain the other measurements separately. The source describes an F1 measure, but its distinct Score and F1 fields are not treated as interchangeable. These archived version 1 protocols remain separate from the current version 2 boards. The complete official download includes configurations hidden by default source filters. Source identities are preserved, and these measurements do not enter overall or category rankings.
Freshness and provenance
Version
FACTS Search V1 2025
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 FACTS Search V1 measure?
The archived FACTS Search V1 protocol measures factual answering with Brave Search and a Gemini 2.0 Flash grader. Its scores remain separate from Search V2.
Which source row has the highest reported FACTS Search V1 value?
Gemini 3.1 Pro Preview has the highest reported FACTS Search V1 value at 85.6%. These archived version 1 protocols remain separate from the current version 2 boards. The complete official download includes configurations hidden by default source filters. Source identities are preserved, and these measurements do not enter overall or category rankings. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on FACTS Search V1?
The Archived version 1 snapshot contains 31 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 6,100+ subscribers.
One email each week. Unsubscribe anytime.