MiniCPM-SALA — benchmark results

OpenBMB's 9B model hybridizing sparse and linear attention for million-token context on consumer GPUs (February 2026). Provider: OpenBMB. Released 2026-02-11. Access: Open.

Unified ELO 1571 ± 17, rank #564 of 1776 rated models, from 10 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELMET (128K)45.4Average Score83.1
LLM Stats (AIME 2024)83.75Score (%)64.2
OpenCompass Research - AIME 202581Score (%)42.6
OpenCompass Research - HLE9.3Score (%)34
OpenCompass Research - LiveCodeBench v654.7Score (%)29.8
OpenCompass58.7Average Score21.3
LLM Stats (CMMLU)81.55Score (%)20
OpenCompass Research - MMLU-Pro72Score (%)8.5
OpenCompass Research - GPQA Diamond58.2Score (%)4.3
OpenCompass Research - IFEval76.9Score (%)2.1

Interactive version: theaggregate.ai/model?slug=minicpm-sala · How the rankings work · Data refreshed daily, snapshot 2026-07-22.