Capability matrix
- Compute and format BF16 · x86 BF16 reference/runtime lane.
- Local entry point Specialized text runner.
- Server Recorded chat pass. Plain chat and streaming on a 4K bundle.
- Harness OpenHands proposed. This variant has no tool protocol. Rebuild and retest prompt admission.
Choose the download
Publisher model card · CKE input files
Publisher task: text generation. CKE example: BF16. Use this artifact identifier:
hf://amd/Instella-MoE-16B-A3B-Think.
Get started
BF16 SafeTensors lane; use the exact command in the v8 runbook Use the model-specific setup.
What has actually run?
Top-1 PyTorch match and 0.99998 full-logit cosine at the tested checkpoint. 32-token checkpoint.
CPU evidence. x86 BF16 reference/runtime lane; quantized and long trajectories open.
Performance. The cited model result does not give a comparable prefill/decode pair here. ARM throughput for this exact artifact is unverified here. Benchmark evidence · 32K to 128K results
Can I fine-tune it?
Full-checkpoint training, LoRA and other PEFT methods: no verified CKE recipe for this exact checkpoint in the reviewed evidence. For a trainable starter, use the FP32 and adapter recipes.