April 23, 2026
By Karan Singh
Tesla has just begun rolling out the latest iteration of its autonomous driving software, Full Self-Driving version 14.3.2. While point updates often focus on minor bug fixes and incremental polishing, this specific release introduces a massive under-the-hood architectural change that fundamentally alters how the FSD’s neural networks operate.
According to the official release notes, the most significant item in the v14.3.2 changelog states that Tesla has officially “unified the model between Actually Smart Summon, FSD, and Robotaxi for more capable and reliable behavior.”
This change applies to Model S/X vehicles as well as Model 3/Y, but unfortunately, leaves out the Cybertruck.
Disengagement Menu
The v14.3.2 changelog also includes a second crucial line: “Help Tesla improve Self-Driving by selecting an intervention reason on the main screen after taking over.”
This new on-screen prompt allows drivers to select from a predefined list of common issues following an intervention or manual takeover. Serving as a faster, tap-to-submit alternative to recording a voice note upon disengagement, this UI addition is likely very similar to the internal feedback tools currently used by Tesla’s QA engineering fleets when testing Supervised Robotaxis. Drivers can still leave a voice note if they’d like, in addition to selecting why they disengaged.
By streamlining why drivers report errors, Tesla can categorize and feed edge-case data back into the training network much more efficiently.
One Unified Model
Previously, different autonomous functions within the vehicle relied on slightly different neural net architectures or segmented training models depending on the task at hand. Navigating a complex, low-speed parking lot using Actually Smart Summon (ASS) required a different operational context than navigating highway traffic on FSD.
By unifying the model, Tesla has effectively combined these disparate systems into a single, cohesive end-to-end neural network. This means the exact same foundational AI that is safely navigating the unsupervised Robotaxi fleets in Dallas and Houston is now handling your morning commute and navigating your vehicle out of a crowded grocery store parking lot.
This brings a massive boost to Actually Smart Summon, which used an older AI model. Early access users are reporting major improvements in summon, with it not only handling more complex situations but also activating much more quickly.
One of the biggest improvements with Actually Smart Summon is the connection/responsiveness from the phone to the car.
The car immediately responds to your button press, even right after stopping the last session, which it would have a big delay on in the past. pic.twitter.com/E7HyfYEucd
— Dirty Tesla (@DirtyTesLa) April 23, 2026 Uses Robotaxi Model
This consolidation brings several massive advantages to both the end user and Tesla’s engineering team.
With a single unified model, vehicles will no longer have to digitally switch gears when transitioning from a parking lot environment (Summon) to a public road (FSD).
The driving behavior will feel significantly more fluid and consistent from the moment the car is called to the moment it parks at the final destination. That also means that FSD and ASS will be accessible faster, as your vehicle doesn’t need to load and unload different models from memory for similar tasks.
Here’s how quickly Actually Smart Summon starts/reacts on FSD v14.3.2. It’s instant the moment you press start on your phone.
Huge improvement @Tesla_AI! pic.twitter.com/YformXfYWl
— Zack (@BLKMDL3) April 23, 2026
Plus, because all three systems now share the exact same underlying architecture, training data becomes universally applicable. A complex edge case solved by an unsupervised Robotaxi in Texas instantly makes Actually Smart Summon better for a consumer in New York.
For Tesla’s Autopilot engineering team, maintaining a single unified model also reduces software complexity. Instead of patching and updating multiple models, they can focus all of their compute power and engineering resources on improving one master system, accelerating the overall pace of future V14 and V15 updates.
As Tesla continues to rapidly iterate on the V14 architecture, this unification proves that the line between consumer driver assistance and fully autonomous Robotaxi capability is officially blurring.
Bringing Summon to the Cybertruck
The Cybertruck still doesn’t have access to Actually Smart Summon and doesn’t include this note in the FSD v14.3.2 release notes. Based on the release notes alone, it’s not clear whether the Cybertruck is using the new unified model from Robotaxi, but there’s a chance it is, and it simply doesn’t include Actually Smart Summon yet.
This could be the update that future versions build on, eventually adding A.S.S. to the Cybertruck, a feature that has notably been missing since the vehicle launched two years ago.
Customers have been waiting for this particular feature for Tesla’s now-flagship vehicle for a long time, and this was one of the key steps to finally enable it.
Subscribe to our newsletter to stay up to date on the latest Tesla news, upcoming features and software updates.
April 23, 2026
By Nehal Malik
Tesla is officially clearing the air regarding the future of its self-driving hardware. During the Q1 2026 Earnings Call on Wednesday, CEO Elon Musk confirmed that the upcoming Full Self-Driving (FSD) v15 will be compatible with current AI4 (Hardware 4) computers, putting an end to months of community speculation.
The news serves to allay growing fears that the next generational leap in software might require Tesla’s upcoming AI5 hardware. Many enthusiasts were concerned that the current AI4 suite, which features a relatively modest 16GB of memory, wouldn’t have the “headroom” to run the massive new models Tesla has in development. However, Musk’s comments confirm that the current fleet is still very much in the race for total autonomy.
Software Overhaul And a 10-Billion-Parameter Model
The anxiety around hardware compatibility wasn’t unfounded. Back in August 2025, Musk teased a massive “10x parameter” upgrade, moving from a 1-billion-parameter model to a significantly more complex 10-billion-parameter version. While this was originally slated to debut late last year with FSD v14, Tesla recently shifted that model upgrade to FSD v15.
Because AI5 chips are expected to feature a drastic increase in onboard memory, many assumed the delay was because the 10-billion-parameter model simply wouldn’t fit on current cars. Musk has now put that theory to rest. While AI5 will eventually offer more power for Tesla’s engineers to utilize, the software overhaul coming with v15 is being designed to squeeze every ounce of performance out of the existing Hardware 4 units.
Safer Than a Human: The Road to v15
During the call, Musk was confident about the current state of the software, noting that Tesla’s vehicles are all autonomy-ready, and v14 is already safer than human drivers. The recently released FSD v14.3 has already shown impressive smoothness and faster reactions in real-world testing, and each point release continues to bring Tesla owners closer to unsupervised autonomy.
However, v15 represents a total architectural overhaul intended to take safety to another level — orders of magnitude beyond human capability. Musk said that Unsupervised FSD will debut on a version of v14 or v15 where it is legal, but Tesla wants to finish training and validating v15 before pushing for larger-scale Unsupervised FSD and Robotaxi deployments. FSD v15 is targeted to launch at the end of 2026 or the start of 2027, making it the most significant software milestone since the move to end-to-end neural networks.
A Future-Proof Fleet
The confirmation that v15 will run on HW4 is a relief for the millions of owners who have recently taken delivery of a new Model 3, Model Y, or Cybertruck. It proves that Tesla’s strategy of “distillation” — shrinking complex AI models to fit on automotive-grade hardware — is still bearing fruit.
As Tesla prepares for mass production of AI5 next year, the current fleet remains the primary focus for achieving unsupervised autonomy. With v15 promised by early next year at the latest, the “autonomy-ready” promise Tesla has made for a decade is finally entering its final, most critical phase.
April 22, 2026
By Karan Singh
Today, on Earth Day, Tesla held its 2026 Q1 Earnings Call and spoke to all the big items from the first quarter of the year, and everything that’s coming next.
Elon’s opening statement noted that it will be an exciting year for Tesla, with substantially increased capital expenditures on research and development, new factories, and new capabilities all at the same time.
Whether you missed it because you were busy or want a handy document to refer to, we’ve got the entire Q1 2026 Earnings Call broken down in neatly summarized notes below for you.
FSD & Robotaxi
Paid Robotaxi miles nearly doubled sequentially during Q1
Tesla officially launched Unsupervised Robotaxi in both Dallas and Houston in April
Elon Musk confirmed there has not been a single accident or injury in the unsupervised program to date
Cybercab will eventually replace the Model Y fleet as the highest volume vehicle on the network in the near future
Tesla secured regulatory approval for FSD (Supervised) deployment in the Netherlands
FSD (Supervised) is moving to a subscription-only model, with record net new subscriptions in Q1
Tesla’s vehicles are all autonomy-ready, and V14 is already safer than human drivers
Unsupervised FSD will become available on a version of V14/V15 where it is legal
FSD V15 is targeted for the end of 2026 or the start of 2027, with a total architectural overhaul
V15 will take safety to another level as part of the overhaul
FSD now boasts 1.3m paid customers globally, with the bulk of that growth coming from new subscriptions
Tesla now actively pitches FSD as a core feature and service during vehicle sales
Following the Netherlands approval, Tesla is preparing for an EU-wide rollout in Q2, though regional regulators gate it
New regulatory approvals have also been received in China, with broader approvals for rollout expected in Q3 2026
Limiting factors for Robotaxi and Unsupervised FSD are validation, as Tesla wants to ensure no accidents or injuries occur during rollout
V14.3 launched in April with significant architectural advances, more coming soon
Robotaxi Deployment Status:
State
Metro Area
Status / Target
California
SF Bay Area
Safety Driver (Current)
Texas
Austin
Ramping Unsupervised
Texas
Dallas
Ramping Unsupervised
Texas
Houston
Ramping Unsupervised
Arizona
Phoenix
Preparations Underway
Florida
Miami
Preparations Underway
Florida
Orlando
Preparations Underway
Florida
Tampa
Preparations Underway
Nevada
Las Vegas
Preparations Underway
Tesla AI and Chips
Cortex 2 Supercomputing cluster is now online and running training workloads
Final chip design for AI5 has been completed
Major partnership with SpaceX to establish the TERAFAB, vertically integrating logic, memory, and packaging in one location
Digital Optimus is being developed to automate digital workloads and complement physical, real-world AI
AI5 tape out and design stage is complete, and Elon mentioned that it is the best edge compute inference chip in existence, and the best value for money
There is already significant momentum and engineering work underway for the future AI6 chip and the Dojo 3 supercomputer chip
Construction on a new Research Technology Fab (Research Fab) at Gigafactory Texas, part of the broader TERAFAB project, will begin this year.
Optimus
Preparations for the first large-scale Optimus factory will begin shortly in Q2
First-gen line is designed to produce 1m robots per year, and is replacing the old Model S and Model X lines in Fremont
Gigafactory Texas is being prepared for the second-gen line, designed for long-term production capability of 10 million robots annually
Optimus mass production will begin this year, and is expected to be useful outside of Tesla’s facilities by the summer of 2027
Optimus V3 is already fully functional, though some aesthetic elements are still being finalized
Tesla is hesitant to show off Optimus V3 due to copycat competitors, but expects to unveil later this summer once production starts
Battery, Powertrain, and Manufacturing
Tesla has begun ramping up North American lithium, cathode, and LFP production capabilities
The production ramp includes new LFP cells in Nevada, alongside cathode materials and the new Lithium Refinery in Texas
Total battery pack capacity remains a limiting factor for vehicle production, not demand
Production for both Cybercab and Semi has officially started
Tesla is experiencing a resurgence in demand across France, Germany, South Korea, and Japan, alongside slight growth in the US
Tesla has achieved its highest Q1 order backlog in two years, driven heavily by the compelling value of the Model 3 and Model Y Standard trims
While the recent gas price increases have positively impacted demand, executives noted that order improvements began before fuel prices spiked
Supply chains for new components have a stretched-out S-curve, meaning initial production for new lines (Optimus, Semi, Cybercab) will be slow, but the ramp will be rapid
Tesla Energy
Progress continues at the new Megafactory outside Houston, with Megapack 3 and Megablock production on track to begin later this year
Total energy storage deployed in the first quarter reached 8.8 GWh
Demand for Megapack remains exceptionally strong, but increasing competition and recent tariffs are having an outsized negative impact on profit margins for Tesla Energy
Tesla has begun meaningful customer deployments of its new in-house designed solar panel produced at Gigafactory New York
The newly designed solar panel features 18 individual power zones, which is three times more than a conventional residential panel
Financials
Tesla produced 408,386 total vehicles in Q1 and successfully delivered 358,023
Total quarterly revenue reached $22.4 billion, representing a 16% year-over-year increase
Tesla reported a GAAP operating income of $0.9 billion for the quarter, with a reported free cash flow of $1.4 billion
Total cash and investments increased by $0.7 billion, leaving Tesla with $44.7 billion at the end of the quarter
Financial outflows for the quarter included $2.0 billion for a SpaceX equity investment
Automotive margins, excluding regulatory credits, increased from 17.9% to 19.2%.
Interest rates and tariffs continue to drag on auto margins, and Tesla has not yet realized any additional income from the recent Supreme Court tariff ruling
Forward-Looking Outlook
Volume production for the Cybercab, Tesla Semi, and Megapack 3 remains firmly on schedule to begin in 2026
Tesla expects its traditional hardware-related profits to eventually be accompanied by an acceleration of AI, software, and fleet-based profits over time
The company remains focused on optimizing its existing production capacity before building entirely new factories or production lines
Tesla expects a massive $25 billion in capital expenditures for 2026.
This will fund six factories, AI research and development, infrastructure, TERAFAB, and solar manufacturing equipment
Investor Q&AInstitutional (Submitted)
Q1: What enabled you to finish the AI5 tapeout early and were there any changes to the original vision? Last week Elon said AI5 will go into Optimus and the supercomputer, but one month ago said it would go into the robotaxi. Has AI5 been dropped from the vehicle roadmap?
A1: AI5 tapeout finished early because the team worked incredibly hard to make it happen. Tesla’s teams worked 6 months straight, through holidays and weekends, to achieve this. AI5 will be focused on Optimus and data centers, and Unsupervised FSD will be achieved by AI4 hardware at greater-than-human safety levels.
AI5 is not immediately needed in vehicles, but that is likely several years away, not a pressing issue – the only reason to replace it is when AI4 gets outdated. AI4 will be receiving an upgrade, with newer-gen RAM, going from 16GB to 32GB, with a 10% compute increase and memory bandwidth increase. AI4.5 will begin production next year. Samsung is currently preparing the modifications to begin production.
Q2: Now that FSD has been approved in the Netherlands and is expected to launch across Europe this summer, can you discuss your robotaxi strategy for the region?
A2: Depends on the governments in Europe to decide what to do, how much they will approve. Tesla expects FSD (Supervised) to begin EU review in May, with Supervised rollout after that. Robotaxi and Unsupervised Rollout will be another red-tape challenge, and the time frame for that is very complex due to regulators.
The iteration deployed in the Netherlands is the same FSD as the US or other regions, just trained on additional local data.
Q3: Given the recent NHTSA incident filings, can you update us on the robotaxi safety data? If safety validation remains the primary bottleneck, why not deploy thousands of vehicles to accelerate removal of the safety driver?
A3: QA fleet and Customer fleet are being used to obtain metrics as fast as possible, scaling is happening, but in a safe way. Tesla is solving the scaling issues – like preventing Robotaxi from blocking intersections, or conducting inaccurate drop-offs.
Q4: Is V14.3 still the last piece of the puzzle to enable large-scale Unsupervised FSD and Robotaxi, or do we need to wait till V15?
A4: Elon thinks it is the last piece of the puzzle – but the question is degrees of safety and convenience. Major architectural improvements coming up will improve safety significantly, and it doesn’t make sense for Tesla to deploy at scale until V15 is ready to deploy, due to the major safety improvements.
Tesla will finish training and validating V15 before pushing on larger-scale Unsupervised FSD and Robotaxi deployments, beyond their current planned deployments. V15 will be a major upgrade.
Institutional (Call-In)
Q1: TERAFAB – which party (of Elon’s companies) will take responsibility for which parts of TERAFAB – funding, design, production, licensing, etc.,
A1: Still working out the details of the deployment. For now, the $3B Advanced Chip Fab (research fab) will be built on Giga Texas grounds as a test and trial ground, with a production capacity of a few thousand wafers a month. TERAFAB is aiming to test out new physics, new designs, and see how production works with new technologies.
SpaceX will take care of the initial phase of scaling TERAFAB itself – intercompany work has to be approved by both boards of directors, and that both Tesla and SpaceX shareholders are served well, and there are no conflicts of interest.
Follow-up: What about Intel’s involvement?
Intel is partnering on some of the core manufacturing technologies; the state-of-the-art 14A process is expected to be used, expected to be ready for prime-time by the time TERAFAB is ready to get off the ground.
Memory, logic, and packaging will all be done at the new Advanced Chip Fab to try out new techniques – and likely be a unique facility as no one else does it under one roof or in one facility, enabling recursive R&D.
Q2: If the current install base is 15%, expected to be about 30-35% in North America, is Tesla winning more FSD users than new car sales? Are most subscribers on AI4? Are most drivers on AI4 using FSD? What does success look like?
A2: Ashok agreed – this is the right way to look at it. You can’t look at one quarter over another for churn, but churn has gone down significantly as FSD has gotten better. Subscriptions have also gone up overall, and customers are driving longer on FSD, using it more often, reducing churn.
Follow-up: xAI and Grok integration into Optimus and Teslas?
Optimus will need to be able to work without a cell signal or wi-fi, so compute will need to be on board and needs to work whether it has a connection or not, similar to how FSD currently works. Optimus needs a “manager” or greater intelligence to give it tasks, and the orchestration AI and voice would be Grok. Optimus could work for several hours without management oversight.
Q3: Chip suppliers generally make good money on the chips they sell. The longer term with TERAFAB is to get supply, but how much of TERAFAB is needed to get better economic gains on chips in the medium-term, and when will it begin?
A3: TERAFAB isn’t to generate leverage over chip suppliers – it is a path to get enough chips for all of Tesla and SpaceX’s projects. Industry cannot keep up with the demand, and so they’ve decided to do their own R&D and production.
Q4: FSD can help to drive vehicle sales, and is important – what is Tesla’s view on new vehicle models? Family vehicle? Compact vehicle?
A4: Cybercab is Tesla’s two-person vehicle; most long-term production will be Cybercab. Tesla’s future lineup will be autonomous vehicles of different sizes. Long-term, the only manually driven car will be the Roadster (new debut date – 1 month or so out – validation still ongoing).
Follow-up: Battery Constraints? How to resolve them? In-house or suppliers?
The limiter isn’t the cells, but the pack production capacity itself. Tesla is actively working on resolving this, and more capacity is being added now, with more coming in the future. Berlin just launched the 4680Y, helping with the European demand surge.
Tesla is retooling and upgrading old battery pack facilities to increase and modernize production, and Giga Shanghai is scaling production as well.
Q5: Safety Driver in Austin now gone, Dallas and Houston are coming. What are the key safety metrics for Robotaxi to expand?
A5: Tesla tracks all the varied safety metrics, along with a very large QA fleet spread across the US. That means looking at both real interventions and simulated interventions, Tesla is working to resolve edge cases. Expansions have so far gone as planned.
Robotaxi’s expansion issues are related to convenience issues – Robotaxi is set for maximum safety, and sometimes can get stuck where it doesn’t feel confident enough to continue – a light that never changes from red or a rail crossing with issues.
Elon described a situation where a Waymo crashed into a bus, and Tesla’s Robotaxis were blocking the left turn lane because they could not continue – the challenge is solving issues where the car is not confident enough to work.
Elon described another situation involving infinite loops due to construction around a single entry point, where the Robotaxi will just loop the block and refuse to enter due to safety.
Follow-Up: Sun Glare, Direct Photon Counting, hasn’t received an update on the NHTSA filing
Updated cameras to prevent sun glare have been in production for some time, but NHTSA is asking for a lot of information on older vehicles, and the investigation is expected to be completed sometime soon, positively. Recent software builds have made sure FSD is safer when there is sun glare or build-up on the windshield and prevent FSD from activating if it cannot see at all.
Retail
Q1: When will we have the Optimus v3 reveal? When will Optimus production start since we ended the Model X and S production earlier than mid-year? What’s the expected Optimus production rate exiting this year? What are the initial targeted skills?
A1: Elon opened by mentioning the copycats once again – unveil will be closer to production, in the late July-August timeframe. Last S/X production will be in Early May; the line is being dismantled part-by-part, with final assembly being last.
Not sure what the production rate of Optimus will be, Tesla is building the plane as they fly it – new and unique parts and ramping production of all these things takes time.
Q2: What milestones are you targeting for unsupervised FSD and Robotaxi expansion beyond Austin this year, and how will that drive recurring revenue?
A2: Rollout is very cautious, no injuries or fatalities to date, and Tesla wants to keep it that way.
Q3: When you do expect FSD Unsupervised to reach customer cars?
A3: Elon said he is just guessing – Q4 2026, did not sound very confident. Tesla has to deal with edge cases – unsafe intersections, bad road markings, high-incident zones,
Q4: How will hardware 3 cars reach unsupervised FSD?
A4: HW3 does not have the capability to reach Unsupervised FSD – for the first time, Tesla’s executives have admitted this. For customers who have purchased FSD, a discounted trade-in for AI4 hardware or a free car computer + camera replacement. Tesla will set up conversion facilities in major cities to conduct the process, as it requires partial vehicle disassembly.
Elon also mentions it makes sense to convert HW3 to AI4 to have older cars join the Robotaxi fleet. Ashok jumped in that V14-lite will be coming to HW3 vehicles soon, and will have all the V14 features that are on AI4, coming late June.
Q5: What is Tesla doing to scale the energy generation business with solar? Residential roof deployments have stalled. Will Tesla move to regional solar and battery farms perhaps coupled to superchargers? Deploy solar through utilities?
A5: Strong demand is still shaping up in the second half of the year, even as the residential home solar credit has dropped. Tesla’s own lease-to-own system has made good progress, along with their new solar panel and mounting system. Both utility-scale and residential solar will continue scaling.
Listen to Earnings Call
You can listen to Tesla’s Earnings Call below: