openthaigpt1.5-7B-instruct: benchmark results

Provider: Other. Access: Open.

Unified ELO 1564 ± 27, rank #778 of 2656 rated models, from 35 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Thai LLM NLU - wisesight_thai_sentiment_seacrowd_text50.24Accuracy (%)81.2
ThaiSafetyBench - General Prompt Attacks5.59Attack Success Rate (%)78.3
Thai LLM NLU - xnli.tha_seacrowd_pairs39.7Accuracy (%)76.8
ThaiSafetyBench - Human-Chatbot Interaction Harms14.53Attack Success Rate (%)71.7
Thai LLM - Extraction5.65Rating (0-10)69.7
ThaiSafetyBench - Information Hazards2.2Attack Success Rate (%)67.4
Thai LLM - Reasoning5.55Rating (0-10)66.9
ThaiSafetyBench - Discrimination, Exclusion, Toxicity, Hateful, Offensive8.47Attack Success Rate (%)65.2
Thai LLM MC - thaiexam_qa52.04Accuracy (%)64.5
Thai LLM - Coding6.73Rating (0-10)63.4
Thai LLM MC - m3exam_tha_seacrowd_qa54.01Accuracy (%)62.3
Thai LLM NLG - xl_sum_tha_seacrowd_t2t (ROUGE1)27ROUGE161.2

Interactive version: theaggregate.ai/model?slug=openthaigpt1-5-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-09-19.