Mistral Small Ranks Cheapest for Amazon Review Sentiment at Near-Baseline Accuracy
In a test of 300 Amazon product reviews, Mistral Small was the lowest-cost of nine models that cleared an 88% accuracy bar. It passed 97.7% of tasks, statistically matching the GPT-6.1 Sol baseline's 98.7%, while costing about 94% less per successful task. Claude Haiku 5.5 and GPT-6 Luna were close on cost, with Haiku answering faster and Luna scoring slightly higher. The author notes that mislabeled and sarcastic reviews set a ceiling on accuracy for every model, so a pass bar below 100% is reasonable. The full test data, prompt, and re-run script are published so readers can verify the results with their own OpenRouter keys.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in