- Organizations
- OpenBMB
- MiniCPM-SALA
MiniCPM-SALA: API Pricing, Context Window & Benchmarks
MiniCPM-SALA is a language model from OpenBMB, released in February 2026.
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%
MiniCPM-SALA benchmarks
Capability tiers
Standing within each category, adjusted for leaderboard depth.
Real tasks performance
High-confidence performance for MiniCPM-SALA across real-world prompt categories. Only 95% intervals at most 4 points wide are shown.
Performance by conversation depth
How MiniCPM-SALA holds up as conversations get longer.
Quality Tracker
MiniCPM-SALA Performance Across Datasets
Scores sourced from the model's scorecard, paper, or official blog posts
MiniCPM-SALA model size
MiniCPM-SALA has 9.5 billion parameters. See how it compares to other models in the same parameter range.
MiniCPM-SALA latency
MiniCPM-SALA time to first token, sustained output throughput, and failed-request rate from live API traffic over the trailing 7 days.
MiniCPM-SALA examples
Recent arena outputs from MiniCPM-SALA, picked from the highest-ranked matchups.
MiniCPM-SALA license
MiniCPM-SALA is released under the Apache 2.0 license, which permits commercial use, has 9.5B parameters.
- License
- Apache 2.0
- Commercial use allowed
- Parameters
- 9.5B
Apache License 2.0 - allows commercial use
MiniCPM-SALA resources
Official sources for MiniCPM-SALA: paper or system card, source repository, model weights.
MiniCPM-SALA vs other models
The most-compared alternatives to MiniCPM-SALA are Kimi K2 0905, Phi 4 Reasoning Plus, Claude 3.5 Sonnet. Open any pair side-by-side for benchmarks, pricing, context, and latency.
Models like MiniCPM-SALA
Models ranked just above and below MiniCPM-SALA by LLM Stats score.
FAQ
Common questions about MiniCPM-SALA.