XPENG Confirms Tesla FSD Rival 'VLA 2.0' for Australia
Xpeng says Australia will be among the first global markets to receive its VLA 2.0 AI driving system from 2027, led by triple Turing chip hardware.
Xpeng will bring its next-generation VLA 2.0 artificial intelligence driving technology to Australia from 2027, with local buyers set to be among the first outside China to receive the system.
The advanced software rollout forms a core pillar of Xpeng's aggressive new right-hand-drive strategy, which will see the brand lock in every future model for Australia, spearheaded by the local arrival of the G9L and L03.
Alex Tang, Head of the International Business Centre at XPENG, confirmed the timing during a media roundtable at the company’s Guangzhou headquarters, where he said Australia is one of the brand’s strategic markets.
"2027 will be the first year that we roll out VLA 2.0 into the global market," Tang said.
"For Australia, of course, [it] will be the first batch of the market that we are trying to implement."
The rollout will centre on Xpeng’s Vision Language to Action system, a vision-led AI platform designed to learn in a more human-like way and reduce information loss between what the vehicle sees, interprets and does.
Xpeng says the technology is intended to deliver higher inference efficiency and quicker vehicle responses, forming part of a broader physical AI strategy spanning passenger cars, robotaxis, humanoid robots and flying cars.
For road cars, the key hardware change is the move to three Xpeng Turing chips in new global models, delivering a combined 2250 TOPS of compute.
Tang said newly launched global models will use the triple-chip layout from the L03 onwards, with that hardware required to run the full VLA 2.0 large model.
The L03 is already on Xpeng’s global product roadmap, with CarSauce recently covering the Mona L03 SUV ahead of its China debut.
| Item | Detail |
|---|---|
| Global rollout timing | Starts in 2027 |
| Australia timing | First batch of global markets, subject to regulations |
| Core system | VLA 2.0, or Vision Language to Action |
| New global model hardware | Three Xpeng Turing chips from the L03 onwards |
| Combined compute | 2250 TOPS |
| China legacy hardware pathway | Model distillation for one or two-chip vehicles, targeting about 80 per cent capability |
| China timing for distilled model | Targeting Q3 2026 |
| Overseas legacy hardware status | Not yet committed due to local regulatory requirements |
For existing and lower-tier vehicles, Xpeng is developing a separate pathway in China using what Tang described as model distillation.
That process reduces the size of the AI model so it can run on vehicles with one or two chips, rather than the full three-chip configuration.
Tang said the smaller model could achieve about 80 per cent of the larger model’s capability, with a China rollout targeted for the third quarter of 2026.
However, he stopped short of promising the same approach for overseas markets, including Australia.
"In the overseas market, because right now we have not even [rolled out] the bigger model, we cannot promise that smaller model," Tang said.
"If it is possible, we will try to bring that information."
The distinction is important for Australian Xpeng buyers because the company is preparing a broader local product push while existing G6 owners remain on earlier hardware.
Tang said current owners would continue to be supported, including through service and software updates, as the brand shifts to a direct Australian operation.
That follows a turbulent period for the brand locally, amid its TrueEV court dispute.
Xpeng is also planning more local validation work for future models sold in Australia.
"For Australia, of course we will do a lot of local tests," Tang said.
He said local development would cover vehicle behaviour and driver assistance software, with AI allowing faster adjustment than traditional hardware-led tuning alone.
More Australian details are due at Xpeng’s local brand “launch” event in Melbourne on 21 July.
Xpeng VLA 2.0 Australia FAQs
Xpeng says 2027 will be the first year for the global rollout of VLA 2.0, with Australia included in the first batch of markets.
New global Xpeng models starting with the L03 are set to use three Xpeng Turing chips with a combined 2250 TOPS of compute.
Xpeng has only committed to VLA 2.0 on vehicles fitted with the triple Turing chip architecture. It has not yet committed to overseas backward compatibility for older hardware.
Xpeng says model distillation shrinks a larger AI model so it can run on vehicles with fewer chips, targeting about 80 per cent capability in China.
Yes. Alex Tang said Xpeng will conduct local testing in Australia, including for driver assistance systems.



