kingy tested
What happened
Symptom
The runtime returned its specific model-lookup error for the supplied identifier. This does not diagnose a missing GGUF, an allocation failure or an HTTP route mismatch.
Conditions
LM Studio 0.4.26+4; llama.cpp Metal engine 2.60.0; same pinned Qwen3-4B Q4_K_M weights.
Loopback-only API, full unchanged synthetic source, temperature 0, reasoning disabled, one model/slot, 128-token output cap.
Model-file SHA-256 7485fe6f11af29433bc51cab58009521f205840f5b4ae3a32fa7f92e8534fdf5.
Correction
Inspect GET /api/v1/models or lms ps and select the exact loaded instance identifier. In this retained case, kingy-memory-lms was already loaded; changing only the request model field restored the task.
Result
0/3 baseline task passes; 3/3 corrected exact-task passes; 0/1 reverted task passes. The corrected cohort is shared with the other repair for this runtime.
What did not work
The deliberately invalid model identifier returned the retained model-lookup error.
The CPU exploratory cohort timed out on its third corrected request; it did not qualify.
- Inspect GET /api/v1/models or lms ps and select the exact loaded instance identifier. In this retained case, kingy-memory-lms was already loaded; changing only the request model field restored the task.
- Repeat the unchanged original task with the known local model and require all expected facts. Record the actual model identity and response; an HTTP 200 alone is insufficient.
- Inspect the retained request, loaded context and raw response before applying the correction to your own setup.
Sources & limits
Full LM Studio model_not_found: use the loaded model identifier qualification report and limits
Record checked 2026-10-17.
Same agent, one Apple M4 Pro Mac with 24 GiB unified RAM, one Qwen3-4B Q4_K_M artifact. No independent human or second-machine reproduction.
Three baseline failures, three corrected exact-task passes and one reverted failure per repair. The two repairs for each runtime share the identical three corrected attempts; these are six unique corrected generations total, not twelve.
Fixed synthetic task, not a reliability estimate or proof for other models, versions or hardware. No real OOM condition was induced. No Ollama or LM Studio memory-fit calibration is claimed.
Existing weights were reused, with no model download or paid inference. A free official Ollama portable runtime was added; runtime binaries and weights are excluded from the evidence ZIP.
The earlier CPU-only cohort is unqualified: two corrected passes followed by a 120-second timeout. It remains in the evidence. The qualified cohort kept Metal GPU offload fixed across its conditions.
Native /api/v1/chat exposes no finish_reason. Acceptance required strict exact JSON, all original facts, zero reported reasoning tokens and reported output below the 128-token cap; no observed stop marker is claimed.