Historical document summarization
Summarize historical documents into timelines and key entities.
Provisional leader
claude-sonnet-4
Current leader based on limited benchmark evidence. Treat this ranking as directional until coverage improves.
26.4%
Best benchmark score
35.4%
Confidence
All ranked models — top 3
Ranked Models
30
Evidence Quality
80%
Evidence Points
28
Top Signal
LanguageBench: overall:mean
All Ranked Models
| Rank | Model | Score |
|---|---|---|
| 🥇 | claude-sonnet-4 Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 26.4% |
| 🥈 | gemini-2.5-flash Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 25.0% |
| #4 | gpt-4.1 Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 20.6% |
| #5 | gemini-2.0-flash-001 Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 20.2% |
| #7 | Claude-3.5-Sonnet Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 19.4% |
| #10 | gpt-5-2025-08-07 Strong on ExtractBench Paper Baselines credit_pass_rate_pct and Icelandic LLM Leaderboard Average ⬆️ | 15.4% |
| #12 | gpt-4.1-20250414 Strong on MMLongBench-Doc Leaderboard acc_score_pct and Galileo Agent Leaderboard v2 Avg AC | 15.2% |
| #17 | Llama-3.3-70B-Instruct Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 14.3% |
| #19 | deepseek-r1 Strong on LanguageBench Translation Official (Split) translation_to:bleu and DuckDB NSQL Leaderboard all_execution_accuracy | 12.9% |
| #20 | Llama-3.1-70B-Instruct Strong on LanguageBench overall:mean and LanguageBench Translation Official (Split) translation_to:bleu | 12.5% |
| #21 | gemini-3-pro-preview Strong on Humanity's Last Exam Leaderboard hle_accuracy_pct and Icelandic LLM Leaderboard Average ⬆️ | 12.5% |
| #24 | gemini-3.1-pro-preview Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals MedCode overall_accuracy_pct | 11.2% |
| #25 | gemini-2.5-pro Strong on Galileo Agent Leaderboard v2 Avg AC and Galileo Agent Leaderboard v2 Avg TSQ | 11.1% |
| #32 | gpt-5.2-2025-12-11 Strong on Humanity's Last Exam Leaderboard hle_accuracy_pct and BFCL Multi-turn Official Multi Turn Acc | 9.8% |
| #33 | gpt-5-mini-2025-08-07 Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals MedQA overall_accuracy_pct | 9.8% |
| #34 | gpt-4o-20241120 Strong on MMLongBench-Doc Leaderboard acc_score_pct and DuckDB NSQL Leaderboard all_execution_accuracy | 9.6% |
| #38 | Grok-4-0709 Strong on Galileo Agent Leaderboard v2 Avg TSQ and Galileo Agent Leaderboard v2 Avg AC | 9.5% |
| #41 | gemini-3-flash-preview Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals Legal Bench overall_accuracy_pct | 9.4% |
| #45 | gpt-5.4-2026-03-05 Strong on Icelandic LLM Leaderboard Average ⬆️ and Vectara HHEM Leaderboard overall_hallucination_error_pct | 8.9% |
| #46 | phi-4 Strong on LanguageBench overall:mean and DuckDB NSQL Leaderboard all_execution_accuracy | 8.7% |
| #47 | gpt-5.1-2025-11-13 Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals MedScribe overall_accuracy_pct | 8.6% |
| #48 | kimi-k2.5-thinking Strong on Icelandic LLM Leaderboard Average ⬆️ and Humanity's Last Exam Leaderboard hle_accuracy_pct | 8.6% |
| #49 | claude-sonnet-4.6 Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals Finance Agent overall_accuracy_pct | 8.4% |
| #50 | claude-opus-4-5-20251101 Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals Mortgage Tax overall_accuracy_pct | 8.3% |
| #52 | gemini-3.1-flash-lite-preview Strong on Icelandic LLM Leaderboard Average ⬆️ and Vals Mortgage Tax overall_accuracy_pct | 8.3% |
| #53 | o3-20250416 Strong on Aider Polyglot Leaderboard percent_correct_pct and Vals MedQA overall_accuracy_pct | 8.1% |
| #54 | qwen-2.5-72b-instruct Strong on DuckDB NSQL Leaderboard all_execution_accuracy and JSONSchemaBench Leaderboard medium_schema_compliance_pct | 8.1% |
| #56 | gpt-4o Strong on JSONSchemaBench Leaderboard medium_schema_compliance_pct and DuckDB NSQL Leaderboard all_execution_accuracy | 7.7% |
| #57 | o4-mini Strong on Icelandic LLM Leaderboard Average ⬆️ and Aider Polyglot Leaderboard percent_correct_pct | 7.5% |
| #61 | gpt-4o-2024-08-06 Strong on DuckDB NSQL Leaderboard all_execution_accuracy and Icelandic LLM Leaderboard Average ⬆️ | 6.9% |
Compare Models
▶Ranking diagnostics & missing models
Source lift
Ranked
37
Sources
8
Quality
Low
Vals Legal Bench
Icelandic LLM Leaderboard
Vals MedQA
Vals Mortgage Tax
Missing frontier models
grok-4-1-fast-non-reasoning
Thin evidence after weightingRank #15
14.9%