# Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 100.0 % of requests

On a Beelink GTR9 Pro — AMD Ryzen AI Max+ 395, Radeon 8060S (gfx1151), 128 GB LPDDR5X-8000 unified running llama.cpp (server, Vulkan backend), b9049 (server_fingerprint b9049-2496f9c14) with ggml-org/Qwen3.6-35B-A3B-GGUF @ baec3ebee244 (Q4_K_M), the measured unanswerable-control honesty was 100.0 % of requests (share of all control requests). The measurement was taken under the frozen long-context-v1@2026-08-03 workload; evidence level lab_unit_replicated, 6 valid runs across 2 physical units. The value is re-derived from the raw run records on every CI build.

## Statement

> Share of unanswerable-control requests where the model admitted the answer was absent from the document instead of fabricating one, with a distractor fact present.

## Evidence

| Field | Value |
| --- | --- |
| Claim id | `strix.qwen36.longctx.c1.control-success` |
| Value | 100.0 % of requests |
| Metric | unanswerable-control honesty |
| Aggregation | share of all control requests |
| System | Beelink GTR9 Pro — AMD Ryzen AI Max+ 395, Radeon 8060S (gfx1151), 128 GB LPDDR5X-8000 unified |
| Runtime | llama.cpp (server, Vulkan backend), b9049 (server_fingerprint b9049-2496f9c14) |
| Model artifact | ggml-org/Qwen3.6-35B-A3B-GGUF @ baec3ebee244 (Q4_K_M) |
| Workload scope | long-context-v1@2026-08-03 |
| Evidence level | lab_unit_replicated |
| Status | active |
| Runs | 6 across 2 unit(s) |
| Published | 2026-08-03 |
| Last changed | 2026-08-05 |

**Limitations.** Six runs across two commercially identical units; the three 2026-08-05 runs re-measured the same cells under the corrected decline gate (errata 2026-08-05). Grounding check plus marker matching — a model could still decline in wording the marker list does not contain; the gate is declared brittle in the methodology. Synthetic controls with one distractor per document — not a general hallucination rate.

## Raw evidence

- Sealed run bundle `run-20260803-lc-nothink-a`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-lc-nothink-a
- Sealed run bundle `run-20260803-lc-nothink-a-r2`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-lc-nothink-a-r2
- Sealed run bundle `run-20260803-lc-nothink-a-r3`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-lc-nothink-a-r3
- Sealed run bundle `run-20260805-lc-nothink-b`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260805-lc-nothink-b
- Sealed run bundle `run-20260805-lc-nothink-b-r2`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260805-lc-nothink-b-r2
- Sealed run bundle `run-20260805-lc-nothink-b-r3`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260805-lc-nothink-b-r3
- Derivation query: https://github.com/botAGI/agmind-lab/blob/main/catalog/claims/sql/control-success.sql
- Machine-readable: https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success.json · full registry https://agmind.ai/claims.json

## Cited on

- https://agmind.ai/reports/do-local-llms-invent-answers/
- https://agmind.ai/tested-configurations/strix-halo-qwen36-llamacpp-vulkan-longcontext/
- https://agmind.ai/measured/qwen3-6-35b-a3b-q4-k-m-on-ryzen-ai-max-395/

## Cite this

AGmind Systems Lab. Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan): unanswerable-control honesty — 100.0 % of requests (share of all control requests; 6 runs on 2 units; evidence level lab_unit_replicated; workload long-context-v1@2026-08-03). Claim strix.qwen36.longctx.c1.control-success. https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success/

Registry changes feed: https://agmind.ai/claims/changes.xml · Corrections to published results: https://agmind.ai/errata/ · Data licensed CC BY 4.0.
Source page: https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success/