The Car G9 Owners Want: Turing Compute at G9 Size, Below G9L Money

A G9 owner charging his car the other night was asked how XPeng's VLA 2.0 driving stack feels. He had not tried it. His car is a dual-Orin G9 with the 8155 cockpit chip — and he is thinking about replacing it, with one specific objection: he does not want a G9L. Too big, he says, and not as good-looking as the G9.

The owner who posted the exchange — 天眼影像 — put it plainly: the G9 experience is good enough that its owner does not want to give it up, but he does want the new assisted driving. The conversation ran so long that two strangers who met at a charger ended up discussing driving, handling and design until well past the poster's usual bedtime.

Photo: 天眼影像 (Weibo).

The two questions put to XPeng

The post ends by asking them of @XP-Candice婷婷, an XPeng staff account, on the owner's behalf:

1. When can a dual-Orin configuration be upgraded to VLA 2.0? And,
2. Can the 6.3.0 interface be reverted to the 6.2.0 interaction design? The poster says he does not want to upgrade to 6.3.0 at all.

The first question has an answer, and it has been the same answer since August. In our coverage of the G9L press conference, He Xiaopeng put the distilled VLA 2.0 build for dual-Orin-X cars in Q4 — the order the community expected all along: single-Turing Max first, dual-Orin X after. Chinese reporting in mid-August framed the same timeline as another delay. The constraint behind it is physical, not commercial: a full 30B VLA 2.0 does not fit the memory bandwidth of a two-Orin-X car, so XPeng is compressing it rather than shipping something unstable — and drivers who have tested the distilled build have come away positive.

The second question has no public answer. XOS 6.3.0 reworked the cockpit's control centre and other interaction surfaces, and XPeng has not said whether owners who prefer the previous layout can keep it.

What we would want, having read that exchange

The G9 owner's objection is not nostalgia, and it is not really about looks. It is that the only way XPeng currently sells you its newest driving stack in a car of that size is to sell you a bigger, more expensive car.

What owners in this position are asking for is specific: a five-seat, G9-sized car with the Ultra/Turing compute and the newer comfort updates, priced below the G9L. Stripped of brand sentiment, that is a request for the software without the square metres.

The pricing is where the argument gets awkward, because something close to that car already exists — and it costs more:

Car	China price (from)
2026 G9 (Turing chip standard across the range)	¥248,800
G9L (list / limited-time launch)	¥241,800 / ¥231,800

So a G9-sized XPeng with the Turing chip is already on sale, and it sits above the G9L on price. The owner's wish — G9 footprint, newest compute, less money than the flagship — is not a gap in the range so much as a gap in the pricing ladder: the smaller car is positioned as the more expensive one, and the complaint from the charging bay is what that does to the people already driving one.

Why XPeng may not build it

There is a commercial logic to the current arrangement. The G9L is the flagship: 5.1 metres, ¥231,800 at launch, and the car XPeng is taking to Paris on 12 October to open European orders. A G9-sized Ultra at a clearly lower price would compete with it directly, and XPeng has just spent a year consolidating product lines precisely to stop models undercutting each other (we covered that reorganisation here).

But the two objections are not symmetric. A dual-Orin G9 owner who wants VLA 2.0 is asking XPeng to finish a software programme it has already announced for Q4 — that is a delivery question. An owner asking for the same software in a smaller, cheaper car is asking for a product decision — and the 2026 G9 refresh, which standardised the Turing chip, shows XPeng already understands the hardware half of it.

Caveats

• The exchange is second-hand. We have the poster's account of the conversation, not the owner's own words, and his car's configuration (dual Orin, 8155) is as reported.
• Q4 is a stated target, not a date. It has already slipped once in the eyes of Chinese coverage; we have no date inside the quarter.
• The interface question is unanswered, and XPeng has made no commitment about keeping the 6.2.0 layout available.
• Prices are China-market and the 2026 G9 figure is a launch price from its own model-year refresh, so the comparison is directional rather than like-for-like on specification.
• The "what we would want" section is our own view, prompted by the exchange — not a report of what XPeng plans.

The bottom line

XPeng's flagship strategy and its installed base are pulling in different directions. The G9L is the right car for the company to be selling in Paris; it is not the car a satisfied G9 owner wants to be told is his only route to the new software. Until there is a G9-sized car with the Turing compute below the flagship's price, every VLA 2.0 release note will generate exactly this conversation — at a charging station, at eleven at night, between two people who would rather not be having it.

Source: 天眼影像 (Weibo), with the 2026 G9 pricing from IT之家 and 第一电动网 — translated & adapted for XPENGTech.