# FACTS Grounding v2

> FACTS Grounding v2 checks whether long-form answers remain supported by supplied documents. Its revised judge ensemble and public/private results remain separate from version 1.

Canonical page: https://benchlm.ai/benchmarks/facts-grounding

- Category: [Reasoning](/reasoning)
- Last updated: October 13, 2026

Evaluation results published by [Google DeepMind, Google Research, and Kaggle](https://www.kaggle.com/benchmarks/google/facts-grounding). undefined

Kaggle labels this benchmark Apache 2.0. We retain aggregate measurements with attribution and source links. That declaration does not establish rights to third-party documents, videos, or other underlying task material; no task contents are included. [Original article terms](https://www.apache.org/licenses/LICENSE-2.0).

FACTS Grounding asks for long-form answers supported by supplied documents of up to 32,000 tokens. Version 2 updates the automated judges. The source describes 1,719 examples across 860 public and 859 held-out cases, and we retain its overall, public, and private scores separately.

Three judges assess factual grounding, with an eligibility check for answers that fail to address the request. The final score averages judge results and the public/private sets. These are document-grounding results under this judge setup; they do not establish open-web factuality or equivalence with version 1.

The complete official download includes models hidden by the source page's default filters. Source names and task versions stay separate from canonical model profiles; these results do not enter overall or category rankings.

These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference.

- [Official Kaggle leaderboard](https://www.kaggle.com/benchmarks/google/facts-grounding)
- [Official leaderboard JSON](https://www.kaggle.com/api/v1/benchmarks/google/facts-grounding/leaderboard)
- [Evaluation method](https://arxiv.org/abs/2512.10791)
- [Benchmark-page Apache 2.0 declaration](https://www.apache.org/licenses/LICENSE-2.0)
- [FACTS suite introduction](https://deepmind.google/blog/facts-benchmark-suite-systematically-evaluating-the-factuality-of-large-language-models/)

## About FACTS Grounding v2

- Year: 2025
- Tasks: 1,719 document-grounded examples: 860 public and 859 held out
- Format: Long-form document-grounded answers
- Paper: [FACTS Grounding v2 methodology](https://arxiv.org/abs/2512.10791)

FACTS Grounding asks for long-form answers supported by supplied documents of up to 32,000 tokens. Version 2 updates the automated judges. The source describes 1,719 examples across 860 public and 859 held-out cases, and we retain its overall, public, and private scores separately. Three judges assess factual grounding, with an eligibility check for answers that fail to address the request. The final score averages judge results and the public/private sets. These are document-grounding results under this judge setup; they do not establish open-web factuality or equivalence with version 1. The complete official download includes models hidden by the source page's default filters. Source names and task versions stay separate from canonical model profiles; these results do not enter overall or category rankings.

FACTS Grounding v2 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (55 source rows)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [GPT-6 Sol](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-09-25 | — | 85.2% |
| 2 | [GPT-6.1 Sol](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-10-13 | — | 82.4% |
| 3 | [GPT-6 Astra](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-09-11 | — | 82.0% |
| 4 | [Gemma 4 26B A4B](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 80.9% |
| 5 | [Gemma 4 31B](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 80.7% |
| 6 | [Gemini 4 Argon](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-10-12 | — | 78.6% |
| 7 | [GPT-5.2](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 76.2% |
| 8 | [GPT-5.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 76.2% |
| 9 | [Gemini 3.7 Flash](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-08-27 | — | 75.5% |
| 10 | [Grok 4.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-07-15 | — | 74.5% |
| 11 | [Gemini 2.5 Pro](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 74.3% |
| 12 | [Gemini 3.5 Flash](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 73.2% |
| 13 | [Gemini 3.8 Flash](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-09-03 | — | 73.1% |
| 14 | [Gemini 3.6 Flash](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-07-22 | — | 72.5% |
| 15 | [Llama 3 - Grounded Language Model](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-05 | — | 71.8% |
| 16 | [GPT-5.6 Terra](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-07-15 | — | 71.2% |
| 17 | [Gemini 3.5 Flash-Lite](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-07-22 | — | 71.2% |
| 18 | [GPT-5.4](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 70.3% |
| 19 | [Gemini 2.5 Flash](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 70.0% |
| 20 | [Claude Opus 4.8](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 69.7% |
| 21 | [GPT-5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 69.6% |
| 22 | [GPT-5.6 Sol](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-07-23 | — | 68.7% |
| 23 | [Claude Opus 5.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-09-24 | — | 68.5% |
| 24 | [Gemini 3.1 Flash-Lite Preview](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 66.5% |
| 25 | [Claude Haiku 5.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2026-10-09 | — | 65.8% |
| 26 | [Gemini 3.1 Pro Preview](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 65.7% |
| 27 | [Claude Opus 4.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 62.1% |
| 28 | [Claude Sonnet 4.5 (thinking)](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-10-01 | — | 61.8% |
| 29 | [Claude Opus 4.7](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 61.7% |
| 30 | [Gemini 3 Flash Preview](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 59.0% |
| 31 | [Claude Sonnet 4.5](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-10-01 | — | 58.7% |
| 32 | [Gemma 3 12B](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 58.3% |
| 33 | [GPT-5 mini](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-11-14 | — | 58.3% |
| 34 | [deepseek-v3.1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 57.4% |
| 35 | [Gemma 3 27B](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 56.4% |
| 36 | [Claude Sonnet 4.6](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 56.2% |
| 37 | [Claude Sonnet 4](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 56.1% |
| 38 | [DeepSeek-R1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 56.0% |
| 39 | [Claude Opus 4.6](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluation date unreported | — | 54.9% |
| 40 | [Claude Opus 4.1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 54.8% |
| 41 | [Grok 4](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 54.7% |
| 42 | [GPT-5.1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-11-17 | — | 50.0% |
| 43 | [Mistral Large 2](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 49.7% |
| 44 | [Qwen 3 235B A22B Thinking](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 48.2% |
| 45 | [GPT-4.1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 45.6% |
| 46 | [Grok 3 Mini](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 45.6% |
| 47 | [Mistral Small 3.1](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-11-13 | — | 43.3% |
| 48 | [grok-4-fast-reasoning](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-11-14 | — | 43.1% |
| 49 | [Qwen 3 235B A22B Instruct 2506](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 40.8% |
| 50 | [Gemma 3 4B](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 38.4% |
| 51 | [o3](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 36.2% |
| 52 | [DeepSeek-V3](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 33.4% |
| 53 | [gpt-oss-120b](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 30.7% |
| 54 | [o4 mini](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-02 | — | 29.3% |
| 55 | [gpt-oss-20b](https://www.kaggle.com/benchmarks/google/facts-grounding) | Kaggle benchmark v2; task revision 3; Evaluated 2025-09-03 | — | 23.1% |

## FAQ

### What does FACTS Grounding v2 measure?

FACTS Grounding v2 checks whether long-form answers remain supported by supplied documents. Its revised judge ensemble and public/private results remain separate from version 1.

### Which source row has the highest reported FACTS Grounding v2 value?

GPT-6 Sol has the highest reported FACTS Grounding v2 value at 85.2%. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference.

### How many models are evaluated on FACTS Grounding v2?

The October 13, 2026 contains 55 AI models.

### Does FACTS Grounding v2 affect BenchLM's overall score?

Not directly. FACTS Grounding v2 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.