MiniCPM-SALA
by OpenBMB
MiniCPM-SALA is a model from OpenBMB, released on February 11, 2026. It ranks #146 of 220 models by Intelligence Score with a score of 14.4.
MiniCPM-SALA (Sparse Attention and Linear Attention) is a 9B hybrid model built from a MiniCPM-4.0 checkpoint via continual training (~2T tokens, 25% of training-from-scratch cost). It interleaves 25% InfLLM-V2 sparse attention and 75% Lightning Attention layers, achieving up to 3.5x inference speed over dense baselines at 256K tokens. With HyPE (Hybrid Positional Encoding) and NoPE in sparse layers, the model extrapolates to 2048K tokens despite a 520K training length, enabling 1M-token inference on consumer GPUs like the RTX 5090.
Key specs
| Provider | OpenBMB |
|---|---|
| Release date | 2026-02-11 |
| License | open source |
| Intelligence Score | 14.4 |
| Coding index | 20.9 |
Quick answers
Is MiniCPM-SALA open source?
Yes. MiniCPM-SALA is an open model released under the open source license.
When was MiniCPM-SALA released?
OpenBMB released MiniCPM-SALA on February 11, 2026.
How does MiniCPM-SALA compare to Claude Opus 5.5 (Reasoning)?
MiniCPM-SALA scores 14.4 on the Intelligence Score versus 60.5 for Claude Opus 5.5 (Reasoning), the current #1.