XPeng x Banma: 30B-class on-device cockpit AI on the 750-TOPS Turing chip
At the 2026 Yunqi Conference, XPeng and Banma (Alibaba/SAIC-backed) unveiled an automotive-grade cockpit solution running Banma's AutoOmni multimodal model — including the new AutoOmni 2.0-23B-A3B MoE — fully on-device on XPeng's 750-TOPS Turing AI chip.
Why it matters
A demonstrated 30B-class model running stably offline inside a car — multi-turn dialogue, multimodal perception and offline knowledge Q&A without cloud — pushing on-device cockpit models an order of magnitude beyond the usual 4B-7B range.
What you could build with it
A fleet-management startup could use the same on-device-first pattern for delivery trucks: run a mid-size multimodal model locally for driver assistance, voice dispatch and offline route Q&A in connectivity dead zones, syncing only weather and news from the cloud.
Does it hold up?
Too early to judge: evidence is a trade-show demonstration and vendor claims; no production vehicle, latency numbers, or independent in-car tests have been published.
Built with XPeng x Banma
- XPENG, Banma unveil automotive-grade AI cockpit solution at 2026 Yunqi ConferenceTopCarNews · Second report on the joint XPeng-Banma unveiling: on-device first, cloud-assisted architecture, the Turing chip's 750 TOPS, and Banma's claim of in-vehicle deployment cooperation with over 60% of automakers.
Learn more
First spotted on article: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.