Model Registry
Capabilities workers can actually run.
The registry connects user intent to available models, precision, deployment footprint, and observed latency.
7 active0 beta146,568 inferences
Registered models
7
7 ready for production workloads
Active footprint
21
Model instances across online nodes
Total inferences
146,568
Verified DeAI Inferences (Live Mesh)
Catalog
Use this as the visible contract between workload requests and worker capabilities.
DeAI-BGE-M3-Embedding
v1.0activeembedding
Size
M3
Precision
INT8
Latency
2299ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
42.7k runs · last used 8/31/2026
DeepSeek-Math-7B-Prover
vV2.0activellm
Size
7B
Precision
INT8
Latency
2434ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
32.5k runs · last used 8/31/2026
Janus-Pro-7B-Multimodal
vPro-1.0activevision
Size
7B
Precision
FP16
Latency
782ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
30.6k runs · last used 8/31/2026
DeepSeek-R1-Distill-Qwen-8B
vDistill-Q4activellm
Size
8B
Precision
INT4
Latency
2325ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
18.9k runs · last used 8/31/2026
DeepSeek-V3-Full-MoE
v3.1activellm
Size
671B-MoE
Precision
INT4
Latency
2620ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
8.8k runs · last used 9/1/2026
DeepSeek-Coder-V2-Lite-16B
vV2-MoEactivellm
Size
16B-MoE
Precision
INT8
Latency
1247ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
7.3k runs · last used 8/31/2026
DeepSpec-Qwen3-Speculative
vSpec-1.0activellm
Size
Draft-3B
Precision
INT4
Latency
1403ms
worker placements3
EU-DE-01 (Frankfurt Genesis)AP-AU-01 (Sydney Master Hub)AP-SG-01 (Singapore Genesis)
5.7k runs · last used 9/1/2026