Inkampani i-SpaceXAI ka-Elon Musk izolo (Julayi 29) yethule i-Grok Voice Think Fast 2.0, okuyimodeli yayo enamandla kakhulu nehlakaniphe kakhulu ye-AI yezwi-kuya-izwi (speech-to-speech).
Mayelana namanani, le modeli ibiza u-0.08 USD ngomzuzu ngamunye womsindo (Qaphela: ngokwezinga lokushintshisana lamanje, cishe u-0.54 yuan wase-China). Ngokohlelo olusemthethweni, mhla zi-5 ku-Agasti imodeli ye-grok-voice-latest izothuthukiswa isuke ku-grok-voice-think-fast-1.0 iye ku-grok-voice-think-fast-2.0.
Le modeli igxile ekuthuthukiseni ukucabanga ngezwi, ukunemba kokulotshiwe, ikhono lokuxoxa kanye nokwethembeka kokusebenzisa amathuluzi. Ngokwedatha yokuhlolwa okuyisisekelo kwezwi-kuya-izwi eshicilelwe yi-Artificial Analysis, i-AA Speech-to-Speech Quality Index yemodeli entsha ingu-82.9%, okungamaphesenti angu-7.2 ngaphezulu kwe-75.7% ye-Grok Voice Think Fast 1.0.
I-SpaceXAI ithe le modeli ingaqedela ukucabanga ngesikhathi esifanayo nalapho ikhuluma, ngaphandle kwesidingo sokwamukela ukubambezeleka okwengeziwe ukuze ithole ikhono lokucabanga. Le nkampani iphinde yathi inguqulo entsha inciphise ama-Token okucabanga adingekayo empendulweni ngayinye; ezindaweni zokukhiqiza, ukubizwa kwamathuluzi kuvame ukuqedwa ngaphambi kokuba i-ejenti iqede ukukhuluma umusho wayo wokuqala.
Ocwaningweni lwangaphakathi oluhlanganisa izinkulungwane zemishwana nezilimi ezingu-24, i-xAI ithe ukusebenza kokulotshiwe kwe-Grok Voice Think Fast 2.0 kuphakeme ngokuphindwe ka-1.5~2.0 kunokwa-Deepgram Nova 3 kanye ne-ElevenLabs Scribe v2, futhi kuphakeme ngokuphindwe ka-1.4 kunenguqulo 1.0. Ezimo ezinomsindo omkhulu ongemuva nokucindezelwa kwezingcingo, i-xAI ithe umehluko phakathi kwayo namamodeli akhethekile okuguqula izwi libe umbhalo ungakhula ufinyelele cishe izikhathi ezingu-10.