TouchstoneGPT-7B-Instruct: benchmark results

Provider: Other.

Aggregate Rating 1331 ± 52, rank #8070 of 10552 rated models, from 14 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
FinReasoning (Deep Insight) - Justification and Causal Depth34.2Rubric score (0-100): a judge score fused from DeepSeek-V3 a5.6
Fin-RATE0.41Macro Accuracy (imported)0
FinReasoning (Data Alignment)18.1Success-weighted score (0-100): the success rate (items whos0
FinReasoning (Data Alignment) - Calculation17.9Success-weighted score (0-100): the success rate (items whos0
FinReasoning (Data Alignment) - Reasoning12.6Success-weighted score (0-100): the success rate (items whos0
FinReasoning (Data Alignment) - Verification23.8Success-weighted score (0-100): the success rate (items whos0
FinReasoning (Deep Insight)33Rubric score (0-100): a judge score fused from DeepSeek-V3 a0
FinReasoning (Deep Insight) - Architectural Richness and Strictness38.3Rubric score (0-100): a judge score fused from DeepSeek-V3 a0
FinReasoning (Deep Insight) - Completeness and Comparative Analysis30.8Rubric score (0-100): a judge score fused from DeepSeek-V3 a0
FinReasoning (Deep Insight) - Factuality and Contextualization28.7Rubric score (0-100): a judge score fused from DeepSeek-V3 a0
FinReasoning (Semantic Consistency)11.8Composite score (0-100): the mean of sentence-level F1 for l0
FinReasoning (Semantic Consistency) - Fact12Composite score (0-100): the mean of sentence-level F1 for l0

Interactive version: theaggregate.ai/model?slug=touchstonegpt-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-16.