Expired60% confidenceBenchmarkExact time
GPT-4-level models can have a Time to First Token (TTFT) of 500–2000ms, while smaller models such as Llama 3.1 8B can achieve a TTFT as low as 100–300ms.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/14/2026
First Seen
Valid until: 10/12/2026(expired)
Sources
状态机语音Agent实战：Pipecat与Vapi的延迟与准确性权衡
redditr/aiagents7/12/2026
Related Claims
Unverified语音Agent典型端到端延迟预算约为STT finalization 100-300ms + LLM TTFT 200-500ms + TTS TTFA 150-250ms + 网络开销50-100ms，总计500-1150ms74% similarUnverified现代云内网 RTT 可低至 0.1–0.5 毫秒，与本地 SSD 的 fsync 延迟处于同一量级72% similarUnverifiedAt FP16 half-precision computation, the RTX 4090 has a theoretical performance of approximately 330 TFLOPS, while the M2 Ultra's GPU delivers approximately 27.2 TFLOPS.71% similarUnverified将模型权重从FP16压缩到INT8或INT4的量化技术，可使消费级GPU如RTX 4090在本地部署时达到30-80 TPS的实用速度71% similarUnverifiedA parameter scale of 30B is large enough to exhibit capability levels close to GPT-3.5, yet small enough to run on high-end consumer graphics cards (such as the NVIDIA RTX 4090) or mid-scale servers.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/503987API
curl https://kongchang.com/api/v1/knowledge/claims/503987MCP
get_claim(id=503987)