OpenBMB model product
MiniCPM SALA
MiniCPM SALA is a 9B hybrid language model from OpenBMB that combines sparse and linear attention for long-context inference.
Updated Sep 8, 2026. Default version: MiniCPM-SALA
LLMBoard Score24.4MiniCPM-SALA
Coverage20%12 benchmark families
Context windowN/ATokens
Official input priceN/AOfficial price unavailable
MiniCPM SALA Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
MiniCPM SALA Benchmark Results
Benchmark scores for MiniCPM-SALA.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkHumanEval | Score95.12% | Rank01 | Participants66 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 128K (2-needle) | Score28.62% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 128K (4-needle) | Score19.62% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 64K (2-needle) | Score29.77% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 64K (4-needle) | Score20.57% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 64K (8-needle) | Score16.56% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkNoLiMa 128K | Score23.86% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkNoLiMa 32K | Score54.54% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkNoLiMa 64K | Score42.95% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkRULER 1000K | Score86.30% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkRULER 128k | Score89.37% | Rank01 | Participants4 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkRULER 2048K | Score81.60% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkRULER 512K | Score87.10% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkRULER 64k | Score92.65% | Rank01 | Participants4 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMRCR 128K (8-needle) | Score10.12% | Rank02 | Participants2 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench v5 | Score60.48% | Rank04 | Participants9 | Percentile62.50% | EvidenceC | Evaluated |
| BenchmarkMBPP | Score89.11% | Rank04 | Participants33 | Percentile90.63% | EvidenceC | Evaluated |
| BenchmarkCMMLU | Score81.55% | Rank05 | Participants6 | Percentile20.00% | EvidenceC | Evaluated |
| BenchmarkBBH | Score81.55% | Rank08 | Participants12 | Percentile36.36% | EvidenceC | Evaluated |
| BenchmarkAIME 2024 | Score83.75% | Rank19 | Participants53 | Percentile65.38% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench v6 | Score52.00% | Rank53 | Participants62 | Percentile14.75% | EvidenceC | Evaluated |
| BenchmarkIFEval | Score76.34% | Rank59 | Participants68 | Percentile13.43% | EvidenceC | Evaluated |
| BenchmarkAIME 2025 | Score78.33% | Rank77 | Participants119 | Percentile35.59% | EvidenceC | Evaluated |
| BenchmarkMMLU-Pro | Score67.04% | Rank103 | Participants138 | Percentile25.55% | EvidenceC | Evaluated |
MiniCPM SALA Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
MiniCPM SALA Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- N/A
- Tracked offerings
- 0
No provider prices
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
MiniCPM SALA Runtime Performance
Provider-specific output speed and catalog latency for MiniCPM-SALA. Runtime does not affect the capability score.
No runtime data
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
MiniCPM SALA Specifications
Technical details for the model's default version.
- Version
- MiniCPM-SALA
- Released
- Feb 11, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 9.5B
- Context window
- N/A
- Max output
- N/A
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- Apache 2.0
MiniCPM SALA Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionMiniCPM-SALA | Released | LLMBoard24.44 | Parameters9.5B | ContextN/A | Max outputN/A | Open weightsYes | LicenseApache 2.0 |
Models similar to MiniCPM SALA
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AC
Qwen2 VL 72B
Alibaba Cloud / Qwen Team
IB
IBM Granite 4.2 3B
IBM
DE
DeepSeek-V3
DeepSeek
AM
Nova Pro
Amazon
AC
QwQ 32B
Alibaba Cloud / Qwen Team
OP
GPT-4.1-mini
OpenAI
What is MiniCPM SALA?
Key information about MiniCPM SALA and its available data.
0 checkpoint through continual training on approximately 2T tokens. It interleaves 25% InfLLM-V2 sparse attention and 75% Lightning Attention layers, with HyPE and NoPE in sparse layers, supporting inference up to 2048K tokens after training at 520K tokens.
Data as of 2026-09-08.
FAQ
Common questions about MiniCPM SALA.
When was MiniCPM SALA released?
MiniCPM SALA's default version was released on Feb 11, 2026.
How much does MiniCPM SALA cost?
No official standard PAYG price is currently available for MiniCPM SALA.
Who created MiniCPM SALA?
MiniCPM SALA was created by OpenBMB.
What is the context window for MiniCPM SALA?
A context window is not available for the default version.
Is MiniCPM SALA open weight?
Yes. The default version is marked as open weight under Apache 2.0.
How many API providers offer MiniCPM SALA?
No provider offering is currently linked to the default version.