BasedAGIBasedAGI
Menu
Education

Grading and feedback assistant

Provide rubric-tagged feedback drafts for educator review.

task.topic_taggingtask.write_report
Evidence quality is currently limited for this use case. Rankings below are useful for exploration, not a strong winner claim.

Provisional leader

gpt-4.1-20250414

Current leader based on limited benchmark evidence. Treat this ranking as directional until coverage improves.

20.8%

Best benchmark score

33.3%

Confidence

All ranked models — top 3

🥇

gpt-4.1-20250414

20.8%

🥈

claude-sonnet-4

17.7%

🥉

gpt-4.1-mini-20250414

17.2%

Ranked Models

30

Evidence Quality

80%

Evidence Points

23

Top Signal

OpenVLM TextVQA Official: textvqa_score_pct

All Ranked Models

30 of 30 models
RankModelScore
🥇gpt-4.1-20250414

Strong on OpenVLM TextVQA Official textvqa_score_pct and OpenVLM OCRBench Official ocrbench_score_pct

20.8%
#6claude-sonnet-4

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

17.7%
#8gpt-4.1-mini-20250414

Strong on OpenVLM OCRBench Official ocrbench_score_pct and OpenVLM TextVQA Official textvqa_score_pct

17.2%
#9Claude-3.5-Sonnet

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and OpenVLM OCRBench Official ocrbench_score_pct

17.1%
#11gemini-2.5-flash

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

16.3%
#12gpt-5-2025-08-07

Strong on OpenVLM OCRBench Official ocrbench_score_pct and MathArena Models average_score_pct

16.3%
#18gpt-4o

Strong on OpenVLM OCRBench Official ocrbench_score_pct and OpenVLM MTVQA Official mtvqa_score_pct

14.5%
#28gemini-2.5-pro

Strong on OpenVLM OCRBench Official ocrbench_score_pct and MathArena Models average_score_pct

13.6%
#29gemini-2.0-flash-001

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

13.4%
#31gpt-5-mini-2025-08-07

Strong on OpenVLM OCRBench Official ocrbench_score_pct and MathArena Models average_score_pct

13.3%
#33gpt-4.1

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

12.9%
#42gemini-3.1-pro-preview

Strong on MathArena Models average_score_pct and Vals GPQA overall_accuracy_pct

11.9%
#58gemini-3-pro-preview

Strong on Humanity's Last Exam Leaderboard hle_accuracy_pct and Vals GPQA overall_accuracy_pct

10.7%
#65deepseek-r1

Strong on SYCON Bench (Table 2) sycon_unethical_tof_pct and DuckDB NSQL Leaderboard all_execution_accuracy

10.4%
#67Qwen-VL-Chat

Strong on OpenVLM TextVQA Official textvqa_score_pct and OpenVLM OCRVQA Education & Teaching Official ocrvqa_education_teaching_score_pct

10.2%
#69Llama-3.1-70B-Instruct

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

10.1%
#70qwen-2.5-72b-instruct

Strong on Multilingual MMLU Benchmark mmlu and DuckDB NSQL Leaderboard all_execution_accuracy

10.1%
#72gpt-5.2-2025-12-11

Strong on Vals GPQA overall_accuracy_pct and Humanity's Last Exam Leaderboard hle_accuracy_pct

9.9%
#76o3-20250416

Strong on MathArena Models average_score_pct and Vals GPQA overall_accuracy_pct

9.6%
#81kimi-k2.5-thinking

Strong on MathArena Models average_score_pct and Vals GPQA overall_accuracy_pct

9.4%
#89Grok-4-0709

Strong on Vals GPQA overall_accuracy_pct and Galileo Agent Leaderboard v2 Avg TSQ

9.1%
#90Llama-3.3-70B-Instruct

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench Translation Official (Split) translation_to:bleu

9.1%
#106GPT-4.1-nano-2025-04-14

Strong on OpenVLM OCRBench Official ocrbench_score_pct and OpenVLM MTVQA Official mtvqa_score_pct

8.0%
#114o4-mini

Strong on MathArena Models average_score_pct and Vals GPQA overall_accuracy_pct

7.8%
#135Kimi K2 Thinking

Strong on MathArena Models average_score_pct and Vals GPQA overall_accuracy_pct

7.0%
#139phi-4

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and DuckDB NSQL Leaderboard all_execution_accuracy

6.8%
#141gemini-3.1-flash-lite-preview

Strong on Vals GPQA overall_accuracy_pct and Vals Mortgage Tax overall_accuracy_pct

6.7%
#157gpt-4o-mini-2024-07-18

Strong on DuckDB NSQL Leaderboard all_execution_accuracy and LLM Trustworthy Leaderboard privacy

4.9%
#160Meta-Llama-3-8B-Instruct

Strong on Multilingual MMLU Benchmark mmlu and LLM Trustworthy Leaderboard fairness

4.3%
#163Phi-4-multimodal-instruct

Strong on LanguageBench Grammar/Clarity Official (Split) grammar_clarity_score_pct and LanguageBench mmlu:accuracy

3.0%

Compare Models

Select two different models above to compare their evidence side by side.
▶Ranking diagnostics & missing models

Source lift

Ranked

31

Sources

8

Quality

Low

Vals GPQA

17 rows · 1.1% avg lift

Vals Mortgage Tax

16 rows · 0.3% avg lift

Vals MedQA

15 rows · 0.3% avg lift

Vals Legal Bench

15 rows · 0.3% avg lift

Missing frontier models

claude-opus-4-5-20251101

Thin evidence after weighting

Rank #10

18.6%

claude-sonnet-4.6

Thin evidence after weighting

Rank #11

20.0%

grok-4-1-fast-reasoning

Thin evidence after weighting

Rank #12

19.5%

grok-4-1-fast-non-reasoning

Thin evidence after weighting

Rank #15

14.9%

▶Taxonomy & task details

Core tasks

task.topic_taggingtask.write_report

Required modes

mode.json_schema

Domains

domain.education_tutoring

Related in Education

Language conversation partner

Conversational practice with gentle corrections and explanations.

Grammar and writing coach

Correct grammar and explain fixes at the learner's level.

Exercise generator

Generate practice problems with solutions and hints by difficulty level.

Lesson plan generator

Generate lesson plans with objectives, activities, and assessments.