# Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token, second question with prompt cache on: 860 ms

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 time to first token, second question with prompt cache on was 860 ms (median over valid requests for this item). The measurement was taken under the frozen doc-session-v1@2026-08-03 workload; evidence level lab_repeated, 3 valid runs across 1 physical unit. The value is re-derived from the raw run records on every CI build.

## Statement

> Median time to first token for the SECOND question over the same 32k-token document with the server prompt cache enabled.

## Evidence

| Field | Value |
| --- | --- |
| Claim id | `strix.qwen36.docsession.c1.ttft-q2-32k-cache` |
| Value | 860 ms |
| Metric | time to first token, second question with prompt cache on |
| Aggregation | median over valid requests for this item |
| 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 | doc-session-v1@2026-08-03 |
| Evidence level | lab_repeated |
| Status | active |
| Runs | 3 across 1 unit(s) |
| Published | 2026-08-03 |
| Last changed | 2026-08-03 |

**Limitations.** Three repeated runs on one unit, reasoning disabled, single stream. Cache benefit requires a byte-identical document prefix in the same server session.

## Raw evidence

- Sealed run bundle `run-20260803-ds-cache-a`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-ds-cache-a
- Sealed run bundle `run-20260803-ds-cache-a-r2`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-ds-cache-a-r2
- Sealed run bundle `run-20260803-ds-cache-a-r3`: https://github.com/botAGI/agmind-lab/tree/main/runs/run-20260803-ds-cache-a-r3
- Derivation query: https://github.com/botAGI/agmind-lab/blob/main/catalog/claims/sql/ttft-ds-en-32k-q2.sql
- Machine-readable: https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q2-32k-cache.json · full registry https://agmind.ai/claims.json

## Cited on

- https://agmind.ai/reports/local-llm-homelab-hardware/
- https://agmind.ai/reports/what-128gb-unified-memory-runs/
- https://agmind.ai/reports/llamacpp-prompt-cache-strix-halo/
- https://agmind.ai/tested-configurations/strix-halo-qwen36-llamacpp-vulkan-docsession/
- https://agmind.ai/compare/prompt-cache-on-vs-off-doc-session/
- 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): time to first token, second question with prompt cache on — 860 ms (median over valid requests for this item; 3 runs on 1 unit; evidence level lab_repeated; workload doc-session-v1@2026-08-03). Claim strix.qwen36.docsession.c1.ttft-q2-32k-cache. https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q2-32k-cache/

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.docsession.c1.ttft-q2-32k-cache/