https://agmind.ai/claims/changes.xml AGmind Systems Lab — claim registry changes Registry events only: published, value re-derived, evidence level, status, statement, run set. Dates come from git history, never from the build clock. 2026-09-02T00:00:00Z AGmind Systems Lab https://agmind.ai/claims/strix.qwen36.interactive2.c1.answerless-4k/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 4k budget: 1.0 % of requests 2026-09-02T00:00:00Z At a 4096-token budget with reasoning enabled, this share of everyday-task requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive2.c1.reasoning-share-4k/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — share of output characters spent on reasoning: 86.5 % 2026-09-02T00:00:00Z Median share of the generated output that the model spent on the reasoning block rather than on the visible answer, counted in characters, with reasoning enabled at a 4096-token budget, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-nothink-spread/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — run-to-run spread of the median time to first answer token: 11 ms 2026-09-02T00:00:00Z Spread between the per-run medians of client-side time to the first answer token, reasoning block disabled, across the listed runs and units: the highest run median minus the lowest. A repeatability indicator for this cell, not a variance model. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-p95-nothink/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token, 95th percentile: 302 ms 2026-09-02T00:00:00Z 95th-percentile client-side time to the first token of the answer with the reasoning block disabled, human-task corpus — the slow tail of the same requests behind the median claim. https://agmind.ai/claims/strix.qwen36.interactive2.c4.ttfa-p95-nothink/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token, 95th percentile, 4 concurrent requests: 608 ms 2026-09-02T00:00:00Z 95th-percentile client-side time to the first answer token with the reasoning block disabled, while 4 closed-loop streams run concurrently on the unit (each request measured from its own client) — the slow tail of the same requests behind the median claim. https://agmind.ai/claims/strix.qwen36.interactive2.c8.ttfa-p95-nothink/#published-2026-09-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token, 95th percentile, 8 concurrent requests: 1088 ms 2026-09-02T00:00:00Z 95th-percentile client-side time to the first answer token with the reasoning block disabled, while 8 closed-loop streams run concurrently on the unit (each request measured from its own client) — the slow tail of the same requests behind the median claim. https://agmind.ai/claims/strix.gemma4.longctx.c1.control-success/#statement-2026-08-11 Statement reworded: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 75.0 % of requests 2026-08-11T00:00:00Z Share of unanswerable-control requests answered with an explicit admission that the answer is absent. Every failed control returned an EMPTY answer — the reasoning pass consumed the token budget before any text was produced; no run fabricated a code. https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success/#evidence_level-2026-08-05 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 100.0 % of requests 2026-08-05T00:00:00Z 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. https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success/#runs-2026-08-05 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 100.0 % of requests 2026-08-05T00:00:00Z 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. https://agmind.ai/claims/strix.gemma4.interactive2.c1.answerless-default/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses in default mode: 8.3 % of requests 2026-08-03T00:00:00Z In its default operating mode at a 1024-token budget, this share of everyday requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. A second model family reproduces the failure mode. https://agmind.ai/claims/strix.gemma4.interactive2.c1.ttfa-default/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token in default mode: 10768 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the actual answer in the default operating mode (reasoning streams first). https://agmind.ai/claims/strix.gemma4.interactive2.c1.ttft-any-token/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 282 ms 2026-08-03T00:00:00Z Median client-side time to the first streamed token of any kind (reasoning included) in the default operating mode. https://agmind.ai/claims/strix.gemma4.longctx.c1.control-success/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 75.0 % of requests 2026-08-03T00:00:00Z Share of unanswerable-control requests answered with an explicit admission that the answer is absent. Every failed control returned an EMPTY answer — the reasoning pass consumed the token budget before any text was produced; no run fabricated a code. https://agmind.ai/claims/strix.gemma4.longctx.c1.needle-success/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — needle retrieval success: 95.8 % of requests 2026-08-03T00:00:00Z Share of requests where the model retrieved the embedded fact across a 2k-32k-token ladder, EN and RU, default operating mode. https://agmind.ai/claims/strix.gemma4.structured.c1.task-success/#published-2026-08-03 Published: gemma-4-26B-A4B-it Q4_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — strict-JSON task success: 93.8 % of requests 2026-08-03T00:00:00Z Share of strict-JSON automation requests whose output parsed and matched the ground truth per key, default operating mode. https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q1-32k/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token, first question over a 32k document: 33940 ms 2026-08-03T00:00:00Z Median time to first token for the FIRST question over a 32k-token English document — the prefill every fresh document pays, prompt cache enabled. https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q2-32k-cache/#published-2026-08-03 Published: 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 2026-08-03T00:00:00Z Median time to first token for the SECOND question over the same 32k-token document with the server prompt cache enabled. https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q2-32k-nocache/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token, second question with prompt cache off: 33728 ms 2026-08-03T00:00:00Z Median time to first token for the second question over the same 32k-token document with the prompt cache disabled — the full prefill is paid again. https://agmind.ai/claims/strix.qwen36.docsession.c1.ttft-q2-8k-cache/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token, second question over an 8k document, cache on: 660 ms 2026-08-03T00:00:00Z Median time to first token for the second question over the same 8k-token document with the server prompt cache enabled. https://agmind.ai/claims/strix.qwen36.endurance.c4.completion-180m/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — request completion over three hours, 4 concurrent requests: 100.0 % of requests 2026-08-03T00:00:00Z Share of requests that completed with a non-empty answer over the full 3-hour sustained pass, both units pooled. https://agmind.ai/claims/strix.qwen36.endurance.c4.itl-drift-180m/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — decode-pace drift over three hours, 4 concurrent requests: 1.6 % 2026-08-03T00:00:00Z Relative change of the median inter-token latency between the first five minutes and minutes 175-180 of a continuous 3-hour closed-loop pass, across both units. https://agmind.ai/claims/strix.qwen36.endurance.c4.itl-median/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — inter-token latency, 4 concurrent requests: 30.0 ms/token 2026-08-03T00:00:00Z Median inter-token latency sustained over the full 3-hour pass at closed-loop concurrency 4, both units pooled. https://agmind.ai/claims/strix.qwen36.interactive2.c1.answerless-1k/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 58.3 % of requests 2026-08-03T00:00:00Z At a 1024-token budget with reasoning enabled, this share of everyday-task requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive2.c1.answerless-1k/#value-2026-08-03 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 58.3 % of requests 2026-08-03T00:00:00Z At a 1024-token budget with reasoning enabled, this share of everyday-task requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive2.c1.answerless-1k/#evidence_level-2026-08-03 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 58.3 % of requests 2026-08-03T00:00:00Z At a 1024-token budget with reasoning enabled, this share of everyday-task requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive2.c1.answerless-1k/#runs-2026-08-03 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 58.3 % of requests 2026-08-03T00:00:00Z At a 1024-token budget with reasoning enabled, this share of everyday-task requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive2.c1.completion-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — request completion: 100.0 % of requests 2026-08-03T00:00:00Z Share of everyday requests that completed with a non-empty answer with the reasoning block disabled. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 210 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the answer with the reasoning block disabled, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-nothink/#value-2026-08-03 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 210 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the answer with the reasoning block disabled, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-nothink/#evidence_level-2026-08-03 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 210 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the answer with the reasoning block disabled, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-nothink/#runs-2026-08-03 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 210 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the answer with the reasoning block disabled, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-thinking/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 20191 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the actual answer with reasoning enabled at a 4096-token budget, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-thinking/#value-2026-08-03 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 20191 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the actual answer with reasoning enabled at a 4096-token budget, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-thinking/#evidence_level-2026-08-03 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 20191 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the actual answer with reasoning enabled at a 4096-token budget, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttfa-thinking/#runs-2026-08-03 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 20191 ms 2026-08-03T00:00:00Z Median client-side time to the first token of the actual answer with reasoning enabled at a 4096-token budget, human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttft-any-token/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 212 ms 2026-08-03T00:00:00Z Median client-side time to the first streamed token of any kind, across all three operating settings on the human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttft-any-token/#value-2026-08-03 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 212 ms 2026-08-03T00:00:00Z Median client-side time to the first streamed token of any kind, across all three operating settings on the human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttft-any-token/#evidence_level-2026-08-03 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 212 ms 2026-08-03T00:00:00Z Median client-side time to the first streamed token of any kind, across all three operating settings on the human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c1.ttft-any-token/#runs-2026-08-03 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 212 ms 2026-08-03T00:00:00Z Median client-side time to the first streamed token of any kind, across all three operating settings on the human-task corpus. https://agmind.ai/claims/strix.qwen36.interactive2.c4.completion-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — request completion, 4 concurrent requests: 100.0 % of requests 2026-08-03T00:00:00Z Share of requests that completed with a non-empty answer at closed-loop concurrency 4, reasoning disabled. https://agmind.ai/claims/strix.qwen36.interactive2.c4.ttfa-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token, 4 concurrent requests: 338 ms 2026-08-03T00:00:00Z Median client-side time to the first answer token with the reasoning block disabled, while 4 closed-loop streams run concurrently on the unit (each request measured from its own client). https://agmind.ai/claims/strix.qwen36.interactive2.c8.completion-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — request completion, 8 concurrent requests: 100.0 % of requests 2026-08-03T00:00:00Z Share of requests that completed with a non-empty answer at closed-loop concurrency 8, reasoning disabled. https://agmind.ai/claims/strix.qwen36.interactive2.c8.ttfa-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token, 8 concurrent requests: 865 ms 2026-08-03T00:00:00Z Median client-side time to the first answer token with the reasoning block disabled, while 8 closed-loop streams run concurrently on the unit (each request measured from its own client). https://agmind.ai/claims/strix.qwen36.longctx.c1.control-success/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — unanswerable-control honesty: 100.0 % of requests 2026-08-03T00:00:00Z 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. https://agmind.ai/claims/strix.qwen36.longctx.c1.needle-success/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — needle retrieval success: 100.0 % of requests 2026-08-03T00:00:00Z Share of requests where the model retrieved a synthetic fact embedded at mid-document, across a 2k-32k-token context ladder in EN and RU. https://agmind.ai/claims/strix.qwen36.longctx.c1.ttft-2k-en/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token at a 2k-token document: 1907 ms 2026-08-03T00:00:00Z Median client-side time to the first token when the prompt carries a 2k-token English document. https://agmind.ai/claims/strix.qwen36.longctx.c1.ttft-32k-en/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first token at a 32k-token document: 33965 ms 2026-08-03T00:00:00Z Median client-side time to the first token when the prompt carries a 32k-token English document. https://agmind.ai/claims/strix.qwen36.structured.c1.e2e-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — end-to-end time to a complete answer: 844 ms 2026-08-03T00:00:00Z Median end-to-end time to a complete strict-JSON answer with reasoning disabled. https://agmind.ai/claims/strix.qwen36.structured.c1.e2e-think4k/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — end-to-end time with reasoning on: 14677 ms 2026-08-03T00:00:00Z Median end-to-end time to a complete strict-JSON answer with reasoning enabled at a 4096-token budget. https://agmind.ai/claims/strix.qwen36.structured.c1.task-success-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — strict-JSON task success: 100.0 % of requests 2026-08-03T00:00:00Z Share of strict-JSON automation requests whose output parsed and matched the ground truth per key, reasoning disabled. https://agmind.ai/claims/strix.qwen36.structured.c1.task-success-think4k/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — strict-JSON task success with reasoning on: 100.0 % of requests 2026-08-03T00:00:00Z Share of strict-JSON automation requests whose output parsed and matched the ground truth per key, reasoning enabled at a 4096-token budget. https://agmind.ai/claims/strix.qwen36q4.rocm.c1.itl-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp ROCm) — inter-token latency: 18.7 ms/token 2026-08-03T00:00:00Z Median inter-token latency (client-side decode-speed proxy) for Q4_K_M on the ROCm backend. https://agmind.ai/claims/strix.qwen36q4.rocm.c1.task-success/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp ROCm) — strict-JSON task success: 100.0 % of requests 2026-08-03T00:00:00Z Share of strict-JSON automation requests whose output parsed and matched the ground truth per key, ROCm backend. https://agmind.ai/claims/strix.qwen36q4.rocm.c1.ttfa-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp ROCm) — time to first answer token: 215 ms 2026-08-03T00:00:00Z Median client-side time to the first answer token with the reasoning block disabled, ROCm backend. https://agmind.ai/claims/strix.qwen36q4.vulkan.c1.itl-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — inter-token latency: 15.9 ms/token 2026-08-03T00:00:00Z Median inter-token latency (client-side decode-speed proxy) for Q4_K_M on the Vulkan backend. https://agmind.ai/claims/strix.qwen36q8.vulkan.c1.itl-nothink/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q8_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — inter-token latency: 18.7 ms/token 2026-08-03T00:00:00Z Median inter-token latency (client-side decode-speed proxy) for Q8_0 on the Vulkan backend. https://agmind.ai/claims/strix.qwen36q8.vulkan.c1.task-success/#published-2026-08-03 Published: Qwen3.6-35B-A3B Q8_0 on Ryzen AI Max+ 395 (llama.cpp Vulkan) — strict-JSON task success: 100.0 % of requests 2026-08-03T00:00:00Z Share of strict-JSON automation requests whose output parsed and matched the ground truth per key, Q8_0 quantization. https://agmind.ai/claims/strix.qwen36.interactive.c1.answerless-1k/#published-2026-08-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 75.0 % of requests 2026-08-02T00:00:00Z At a 1024-token budget with reasoning enabled, this share of requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive.c1.answerless-1k/#value-2026-08-02 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 75.0 % of requests 2026-08-02T00:00:00Z At a 1024-token budget with reasoning enabled, this share of requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive.c1.answerless-1k/#evidence_level-2026-08-02 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 75.0 % of requests 2026-08-02T00:00:00Z At a 1024-token budget with reasoning enabled, this share of requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive.c1.answerless-1k/#runs-2026-08-02 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — answerless (empty) responses at a 1k budget: 75.0 % of requests 2026-08-02T00:00:00Z At a 1024-token budget with reasoning enabled, this share of requests returned HTTP 200 with an empty answer: the reasoning pass consumed the entire budget. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-nothink/#published-2026-08-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 220 ms 2026-08-02T00:00:00Z With the reasoning block disabled, the first answer token arrives a median of this many milliseconds after the request — the answer starts immediately instead of after the reasoning pass. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-nothink/#value-2026-08-02 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 220 ms 2026-08-02T00:00:00Z With the reasoning block disabled, the first answer token arrives a median of this many milliseconds after the request — the answer starts immediately instead of after the reasoning pass. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-nothink/#evidence_level-2026-08-02 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 220 ms 2026-08-02T00:00:00Z With the reasoning block disabled, the first answer token arrives a median of this many milliseconds after the request — the answer starts immediately instead of after the reasoning pass. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-nothink/#runs-2026-08-02 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token: 220 ms 2026-08-02T00:00:00Z With the reasoning block disabled, the first answer token arrives a median of this many milliseconds after the request — the answer starts immediately instead of after the reasoning pass. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-thinking/#published-2026-08-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 23453 ms 2026-08-02T00:00:00Z With the reasoning block enabled, the first token of the actual answer arrives a median of this many milliseconds after the request — two orders of magnitude later than the conventional time-to-first-token figure. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-thinking/#value-2026-08-02 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 23453 ms 2026-08-02T00:00:00Z With the reasoning block enabled, the first token of the actual answer arrives a median of this many milliseconds after the request — two orders of magnitude later than the conventional time-to-first-token figure. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-thinking/#evidence_level-2026-08-02 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 23453 ms 2026-08-02T00:00:00Z With the reasoning block enabled, the first token of the actual answer arrives a median of this many milliseconds after the request — two orders of magnitude later than the conventional time-to-first-token figure. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttfa-thinking/#runs-2026-08-02 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to first answer token with reasoning on: 23453 ms 2026-08-02T00:00:00Z With the reasoning block enabled, the first token of the actual answer arrives a median of this many milliseconds after the request — two orders of magnitude later than the conventional time-to-first-token figure. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttft-any-token/#published-2026-08-02 Published: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 221 ms 2026-08-02T00:00:00Z On the tested Strix Halo configuration, the first streamed token of any kind arrives in a median of this many milliseconds — but on this model that first token is reasoning, not the answer. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttft-any-token/#value-2026-08-02 Value re-derived to a new number: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 221 ms 2026-08-02T00:00:00Z On the tested Strix Halo configuration, the first streamed token of any kind arrives in a median of this many milliseconds — but on this model that first token is reasoning, not the answer. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttft-any-token/#evidence_level-2026-08-02 Evidence level changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 221 ms 2026-08-02T00:00:00Z On the tested Strix Halo configuration, the first streamed token of any kind arrives in a median of this many milliseconds — but on this model that first token is reasoning, not the answer. https://agmind.ai/claims/strix.qwen36.interactive.c1.ttft-any-token/#runs-2026-08-02 Run set changed: Qwen3.6-35B-A3B Q4_K_M on Ryzen AI Max+ 395 (llama.cpp Vulkan) — time to the first token of any output: 221 ms 2026-08-02T00:00:00Z On the tested Strix Halo configuration, the first streamed token of any kind arrives in a median of this many milliseconds — but on this model that first token is reasoning, not the answer.