granite-3.2-8B-instruct — benchmark results

IBM's Granite 3.2 8B instruct model with an experimental toggleable thinking mode for step-by-step reasoning. Provider: IBM. Released 2025-02-26. Access: Open.

Unified ELO 1514 ± 25, rank #754 of 1776 rated models, from 8 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - MuSR16.79Score86.8
Open LLM Leaderboard - IFEval72.75Score86.7
Open LLM Leaderboard - MATH Level 523.79Score77.2
Open LLM Leaderboard - GPQA8.72Score71.9
Open LLM Leaderboard - BBH34.66Score69.1
Open LLM Leaderboard - MMLU-Pro27.91Score51.8
Gorilla API Bench (BFCL)26.87Overall Accuracy (%)25.6
Wolfram LLM Benchmarking Project23Correct Functionality (%)21.7

Interactive version: theaggregate.ai/model?slug=granite-3-2-8b-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-22.