{ "date": "2026-10-29", "model": "qwen3-tts-1.7b-customvoice (Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice)", "runtime": "vLLM-Omni, /v1/audio/speech", "corpus": { "short": "Hello. How can I help you today?", "medium": "Your appointment is confirmed for Thursday at ten in the morning. Please arrive ten minutes early.", "long": "I can help you book an appointment, change an existing booking, or answer a question about our services. Tell me which option you need, and we will take it one step at a time." }, "request": { "model": "qwen3-tts-1.7b-customvoice", "voice": "ryan", "language": "English", "instructions": "Speak warmly and clearly.", "response_format": "wav", "stream": false }, "methodology": { "http": "Sequential, alternating host order per pair; alternating WAV/PCM order per repeat. Same text, Ryan, English and delivery instruction. One persistent HTTP client. No seed, explicit warmup, concurrent benchmark requests or deployment changes. First request retained, not an asserted cold start. HTTP timing is not Chans roundtrip or physical playback. PCM timing assumes mono 24 kHz s16; early silence can precede audible speech. Hardware, utilization, quantization and Omni package revision not inspected. Small samples are not an SLA.", "requests_per_round": 48, "engine_turns_per_host_per_round": 6, "wav_ready": "Complete WAV response received and decoded.", "pcm_first_frame": "First 20 ms of raw 24 kHz PCM received; often near-silent.", "continuous_start_estimate": "Offline max(arrival - preceding audio duration) over each arrival trace; earliest gap-free playback start under ideal playback.", "engine_time_to_first_audio": "Last caller speech frame to first scheduled reply audio in the shared conversation engine, fixed reply text.", "limitations": "GPU identities supplied by the owner, not collected during runs. The R9700 host also served Whisper and the LLM. No GPU utilization, clocks, power, driver or vLLM-Omni build recorded. Reported vLLM version only. Sampling not seeded. 4 samples per text and mode per host per round." }, "rounds": [ { "round": "initial", "started_at_utc": "2026-10-28T23:02:06.676680+00:00", "hosts": { "r9700": { "gpu": "AMD Radeon AI PRO R9700", "reported_vllm_version": "0.27.1", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 2.78936, "completion_s_median": 2.78921, "audio_s_median": 2.16, "rtf_median": 1.27793, "completion_s_max": 3.36111 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 7.85964, "completion_s_median": 7.85937, "audio_s_median": 6.44, "rtf_median": 1.21559, "completion_s_max": 8.18046 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 13.1379, "completion_s_median": 13.13764, "audio_s_median": 11.0, "rtf_median": 1.19474, "completion_s_max": 16.99327 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.16265, "completion_s_median": 2.89962, "audio_s_median": 2.28, "rtf_median": 1.25972, "continuous_playback_start_estimate_s_median": 2.42441, "completion_s_max": 3.50839 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.16397, "completion_s_median": 7.18752, "audio_s_median": 6.0, "rtf_median": 1.19794, "continuous_playback_start_estimate_s_median": 3.10608, "completion_s_max": 9.31446 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.16334, "completion_s_median": 12.11092, "audio_s_median": 10.12, "rtf_median": 1.19157, "continuous_playback_start_estimate_s_median": 3.8102, "completion_s_max": 12.74403 } } }, "engine": { "turns": 6, "response_s_median": 3.1195, "time_to_first_audio_s_median": 3.7255, "endpoint_s_median": 0.605, "stt_s_median": 0.1635, "llm_s_median": 0.6005, "tts_s_median": 2.44, "llm_s_max": 45.444, "time_to_first_audio_s_max": 49.451, "llm_budget_5s_exceeded_turns": 1 } }, "rtx_pro_6000_max_q": { "gpu": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition", "reported_vllm_version": "0.30.0", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.41028, "completion_s_median": 0.41013, "audio_s_median": 1.96, "rtf_median": 0.21235, "completion_s_max": 0.44263 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 1.17399, "completion_s_median": 1.17375, "audio_s_median": 6.16, "rtf_median": 0.19235, "completion_s_max": 1.46603 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 2.20754, "completion_s_median": 2.20718, "audio_s_median": 11.48, "rtf_median": 0.18897, "completion_s_max": 2.6021 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.04285, "completion_s_median": 0.38821, "audio_s_median": 1.88, "rtf_median": 0.20647, "continuous_playback_start_estimate_s_median": 0.30685, "completion_s_max": 0.40994 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.04304, "completion_s_median": 1.0717, "audio_s_median": 5.64, "rtf_median": 0.19001, "continuous_playback_start_estimate_s_median": 0.32206, "completion_s_max": 1.09347 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.04834, "completion_s_median": 2.18691, "audio_s_median": 11.76, "rtf_median": 0.18626, "continuous_playback_start_estimate_s_median": 0.32559, "completion_s_max": 3.01615 } } }, "engine": { "turns": 6, "response_s_median": 0.9455, "time_to_first_audio_s_median": 1.5515, "endpoint_s_median": 0.606, "stt_s_median": 0.165, "llm_s_median": 0.2005, "tts_s_median": 0.419, "llm_s_max": 13.046, "time_to_first_audio_s_max": 14.414, "llm_budget_5s_exceeded_turns": 1 } } } }, { "round": "r9700_upgraded", "started_at_utc": "2026-10-28T23:35:24.595732+00:00", "hosts": { "r9700": { "gpu": "AMD Radeon AI PRO R9700", "reported_vllm_version": "0.32.0", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 1.84374, "completion_s_median": 1.84357, "audio_s_median": 2.08, "rtf_median": 0.93762, "completion_s_max": 2.41326 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 4.58788, "completion_s_median": 4.58754, "audio_s_median": 5.52, "rtf_median": 0.74011, "completion_s_max": 6.23428 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 7.22903, "completion_s_median": 7.22842, "audio_s_median": 12.48, "rtf_median": 0.64348, "completion_s_max": 9.35627 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.0814, "completion_s_median": 0.65173, "audio_s_median": 2.16, "rtf_median": 0.30245, "continuous_playback_start_estimate_s_median": 0.52093, "completion_s_max": 0.71628 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.1478, "completion_s_median": 4.67837, "audio_s_median": 6.48, "rtf_median": 0.69779, "continuous_playback_start_estimate_s_median": 1.54979, "completion_s_max": 4.78633 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.10977, "completion_s_median": 7.49817, "audio_s_median": 11.6, "rtf_median": 0.66206, "continuous_playback_start_estimate_s_median": 1.54551, "completion_s_max": 9.80837 } } }, "engine": { "turns": 6, "response_s_median": 2.3425, "time_to_first_audio_s_median": 2.948, "endpoint_s_median": 0.6055, "stt_s_median": 0.167, "llm_s_median": 0.435, "tts_s_median": 1.8685, "llm_s_max": 0.804, "time_to_first_audio_s_max": 3.912, "llm_budget_5s_exceeded_turns": 0 } }, "rtx_pro_6000_max_q": { "gpu": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition", "reported_vllm_version": "0.30.0", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.42897, "completion_s_median": 0.42882, "audio_s_median": 2.08, "rtf_median": 0.20694, "completion_s_max": 0.44733 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.9911, "completion_s_median": 0.99088, "audio_s_median": 5.2, "rtf_median": 0.19156, "completion_s_max": 1.10963 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 2.01701, "completion_s_median": 2.01651, "audio_s_median": 10.8, "rtf_median": 0.18574, "completion_s_max": 2.26215 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.04253, "completion_s_median": 0.38529, "audio_s_median": 1.88, "rtf_median": 0.20437, "continuous_playback_start_estimate_s_median": 0.30388, "completion_s_max": 0.44449 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.0464, "completion_s_median": 1.17596, "audio_s_median": 6.16, "rtf_median": 0.19098, "continuous_playback_start_estimate_s_median": 0.3339, "completion_s_max": 1.34823 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.04284, "completion_s_median": 2.10647, "audio_s_median": 11.36, "rtf_median": 0.1858, "continuous_playback_start_estimate_s_median": 0.32261, "completion_s_max": 2.62303 } } }, "engine": { "turns": 6, "response_s_median": 0.805, "time_to_first_audio_s_median": 1.4105, "endpoint_s_median": 0.605, "stt_s_median": 0.168, "llm_s_median": 0.185, "tts_s_median": 0.4125, "llm_s_max": 0.565, "time_to_first_audio_s_max": 1.899, "llm_budget_5s_exceeded_turns": 0 } } } }, { "round": "both_upgraded", "started_at_utc": "2026-10-28T23:52:05.472935+00:00", "hosts": { "r9700": { "gpu": "AMD Radeon AI PRO R9700", "reported_vllm_version": "0.32.0", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 1.31407, "completion_s_median": 1.3139, "audio_s_median": 2.16, "rtf_median": 0.62069, "completion_s_max": 1.95751 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 4.11794, "completion_s_median": 4.11768, "audio_s_median": 5.76, "rtf_median": 0.63694, "completion_s_max": 5.09638 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 7.79732, "completion_s_median": 7.79674, "audio_s_median": 11.28, "rtf_median": 0.668, "completion_s_max": 8.92538 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.07922, "completion_s_median": 0.94576, "audio_s_median": 2.0, "rtf_median": 0.44805, "continuous_playback_start_estimate_s_median": 0.71842, "completion_s_max": 1.85405 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.13659, "completion_s_median": 3.48187, "audio_s_median": 4.8, "rtf_median": 0.72658, "continuous_playback_start_estimate_s_median": 0.8907, "completion_s_max": 4.35496 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.10931, "completion_s_median": 6.48838, "audio_s_median": 11.12, "rtf_median": 0.57505, "continuous_playback_start_estimate_s_median": 1.52026, "completion_s_max": 10.58496 } } }, "engine": { "turns": 6, "response_s_median": 2.311, "time_to_first_audio_s_median": 2.917, "endpoint_s_median": 0.606, "stt_s_median": 0.1675, "llm_s_median": 0.2425, "tts_s_median": 1.9215, "llm_s_max": 0.587, "time_to_first_audio_s_max": 3.729, "llm_budget_5s_exceeded_turns": 0 } }, "rtx_pro_6000_max_q": { "gpu": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition", "reported_vllm_version": "0.32.0", "http": { "wav": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.47408, "completion_s_median": 0.47393, "audio_s_median": 2.0, "rtf_median": 0.20525, "completion_s_max": 3.60627 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 1.03499, "completion_s_median": 1.0347, "audio_s_median": 5.48, "rtf_median": 0.18954, "completion_s_max": 1.19454 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 1.9482, "completion_s_median": 1.94775, "audio_s_median": 10.4, "rtf_median": 0.18855, "completion_s_max": 2.11009 } }, "stream": { "short": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.02648, "completion_s_median": 0.43676, "audio_s_median": 2.2, "rtf_median": 0.19871, "continuous_playback_start_estimate_s_median": 0.31502, "completion_s_max": 0.57532 }, "medium": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.0266, "completion_s_median": 1.00219, "audio_s_median": 5.32, "rtf_median": 0.19085, "continuous_playback_start_estimate_s_median": 0.3179, "completion_s_max": 1.12657 }, "long": { "successes": 4, "failures": 0, "first_playable_audio_s_median": 0.02643, "completion_s_median": 1.98866, "audio_s_median": 10.72, "rtf_median": 0.18483, "continuous_playback_start_estimate_s_median": 0.31742, "completion_s_max": 2.09299 } } }, "engine": { "turns": 6, "response_s_median": 0.8435, "time_to_first_audio_s_median": 1.4495, "endpoint_s_median": 0.6055, "stt_s_median": 0.168, "llm_s_median": 0.2425, "tts_s_median": 0.4475, "llm_s_max": 0.613, "time_to_first_audio_s_max": 2.224, "llm_budget_5s_exceeded_turns": 0 } } } } ] }