Capability matrix
- Compute and format Q8_0 preview · eight-thread CPU trial.
- Local entry point Experimental text generation.
- Server Unverified for this artifact. A preview generated 128 tokens; the expected answer was missing.
- Harness No recorded client task. Verify end-to-end output before advertising a server.
Can I use it for chat today?
Experimental. CKE can run a pinned Q8_0 Inkling 2 Mini artifact through text generation. For a first chat session, use the tested Qwen2 command. A recorded Inkling trial generated 128 tokens without emitting the expected answer. There is no recommended everyday chat or fine-tuning command here yet.
What changed in the latest run?
On the same eight CPU threads, processing a 64-token prompt segment after a prepared prefix took about 36 seconds with the newer setting, down from 47 seconds. That timing covers one segment; it doesn't tell you how long an entire conversation takes. What matters next for a user is whether ordinary prompts reliably return useful answers.
Read the answer trial · See the measured update
Files, hardware and training
Publisher model card · GGUF distribution. The tested artifact is Q8_0. No ARM throughput, full-checkpoint or adapter training result is established for that artifact. Long-input checks.