Skip to content
Saved

Expectation vs reality/Robot Vacuums/3 robots compared/Reviewed Jun 2026

What can robot arms actually pick up in homes today?

5 claims tested5 hold only partly.
Roborock Saros Z70View recordRoborockIIISaros Z70
Dreame L50 UltraView recordDreameIIIL50 Ultra
Full specifications25 measurements · 51 of 75 documented
MSRP
$2,599
Not documented
$1,399
Rated runtime
180 min
Not documented
200 min
Charge time
150 min
Not documented
240 min
Battery capacity
92.2 Wh
Not documented
92 Wh
Battery type
Li-ion
Li-ion
Li-ion
Suction power
22,000 Pa
30,000 Pa
19,500 Pa
Noise level
67 dB
Not documented
68 dB
Navigation
StarSight Autonomous System 2.0: dual-transmitter solid-state lidar with 3D time-of-flight array, VertiBeam structured-light lateral sensing, single RGB camera; no rotating lidar turret
TriSight Dynamic System: AstroVision navigation + OmniArm vision (arm-mounted) + StereoEdge laser ToF lateral sensing
LiDAR SLAM with multi-floor mapping, AI camera obstacle avoidance, room segmentation
Connectivity
wifi, bluetooth, alexa, google_home
wifi
wifi, alexa, google_home
Dimensions (W×D×H)
13.8 × 13.9 × 3.1 in
Not documented
13.8 × 13.8 × 3.9 in
Weight
30.4 lbs
Not documented
9.8 lbs
Form factor
Not documented
round
Not documented
Readiness
Ready Now
Promising Progress
Ready Now
Released
May 2025
Sep 2025
Mar 2025
Self-empty
Yes (dock)
Not documented
Not documented
Mop system
Rotating w/ lift
Not documented
Not documented
Robotic arm
Single-object pickup
Not documented
Not documented
Subscription
None required
Not documented
Not documented
Degrees of Freedom
5
Not documented
Not documented
Onboard dustbin
180 mL
Not documented
400 mL
Water tank capacity
69 mL
Not documented
Not documented
Dock type
Auto-empty, hot-water mop wash (176F), warm-air drying, automatic mop removal, detergent dispensing
Auto-empty
Auto-empty, mop washing, hot-water washing, drying
Mop lift
22 mm
Not documented
10.5 mm
Dust bag capacity
2.7 L
Not documented
3.2 L
Threshold climb
1.2 in
2.4 in
1.7 in

Claims under test

5 marketing claims, measured against documented evidence

Each claim below pairs the marketing language with what owner reports, demos, and engineering documentation actually show. The meter summarizes how far each claim sits from the evidence.

5 Partial
Claim 01 / 05

Autonomous clutter pickup before cleaning

Partial

Pickup real but constrained; fragile and irregular objects still need manual clearing

What marketing promises

The robotic arm autonomously picks up floor clutter so you do not have to tidy before each cleaning session.

Roborock Saros Z70 product page and Dreame Cyber 10 Ultra press materials, 2025-2026

What the evidence shows

The Saros Z70's OmniGrip arm retrieves one object per pass and deposits it into a bin. Owner reports and press reviews document success on socks, small toys, and charging cables under 200 g placed in the open. Fragile, irregular, or floor-level-flush objects still require manual clearing before each run.

Z70 arm payload (manufacturer spec)
Under 200 g
Z70 pickup success range (owner reports)
Varies by shape and placement
Objects requiring manual pre-clear
Fragile, irregular, or low-profile
Cyber 10 Ultra: arm validated
Show demo only (June 2026)
L50 Ultra: arm present
None
Claim 02 / 05

Arm handles real-room object variety

Partial

Supported class is narrower than 'everyday clutter'; irregular shapes and flush objects excluded

What marketing promises

The arm handles everyday household objects including cables, toys, socks, and lightweight clutter of varying shapes.

Roborock OmniGrip documentation and Dreame Cyber 10 Ultra CES 2026 press materials

What the evidence shows

The Verge and WIRED reviews of the Saros Z70 note that the arm struggles with non-uniform shapes and objects flush to floor edges. Manufacturer documentation lists socks, toys, and cables under 200 g as the supported class. The Dreame Cyber 10 Ultra's only publicly documented demo showed arm pickup of balls in a structured booth environment; cable and irregular-object handling has not been independently observed.

Z70 documented success objects
Socks, toys, cables under 200 g
Z70 grip force
Fixed; non-adaptive
Cyber 10 Ultra demo object type
Balls (structured demo)
Cyber 10 Ultra real-room cables tested
Not documented
L50 Ultra object handling
No arm; not applicable
Claim 03 / 05

Reduces pre-run floor preparation

Partial

Pre-run prep reduced in predictable clutter scenarios; not eliminated for fragile or irregular objects

What marketing promises

Owning a robot vacuum with an arm means you no longer need to pick up the floor before starting a cleaning session.

Composite from Roborock Saros Z70 and Dreame Cyber 10 Ultra marketing copy, 2025-2026

What the evidence shows

Owner reports across Reddit and product review forums consistently describe a reduced but not eliminated pre-run step. Phones, glasses, ceramics, and unsupported clutter types still require manual removal. The Saros Z70 manual explicitly states fragile items should be pre-cleared. The Dreame Cyber 10 Ultra has no post-shipping owner data as of June 2026.

Z70 manual guidance on fragile items
Recommends pre-clearing
Z70 owner-reported prep reduction
Partial; varies by home type
Cyber 10 Ultra owner data available
None (pre-release)
L50 Ultra pre-run clutter handling
No arm; owner must pre-clear
"No tidying needed" claim status
Not corroborated at scale
Claim 04 / 05

Arm performance validated beyond demo conditions

Partial

Saros Z70 arm ships but draws mixed evidence; Cyber 10 Ultra arm unproven outside demos

What marketing promises

Arm-based manipulation represents a proven, shipping capability ready for regular home use.

Roborock Saros Z70 retail listings from May 2025; Dreame Cyber 10 Ultra described as moving toward commercial readiness at CES 2026

What the evidence shows

The Saros Z70 shipped in May 2025 and has over 12 months of owner data. WIRED's review headline noted the arm 'doesn't quite work'; The New York Times described it as 'mostly succeeds.' The Dreame Cyber 10 Ultra targets an August 2026 ship date with no retail listing or independent test as of June 2026. The L50 Ultra is a proven shipping product with no arm.

Z70 shipping date
May 2025
Z70 WIRED verdict
OmniGrip 'doesn't quite work'
Z70 NYT verdict
'tries and mostly succeeds'
Cyber 10 Ultra ship date
August 2026 (target, unconfirmed)
Cyber 10 Ultra independent test
None as of June 2026
Claim 05 / 05

Premium price reflects arm capability delivered

Partial

Arm delivers real but constrained pickup; $1,200 delta for a conditional, shape-dependent capability

What marketing promises

The premium price over standard flagship vacuums reflects the value of delivered arm functionality.

Implicit in Roborock Saros Z70 ($2,599 MSRP) and Dreame Cyber 10 Ultra (~$2,100 target) positioning vs no-arm L50 Ultra ($1,399 MSRP)

What the evidence shows

The Saros Z70 carries a $1,200 premium over the Dreame L50 Ultra. Press reviews and owner reports document that core navigation and cleaning performance at Level III is comparable across all three robots. The arm accounts for the full price delta; cleaning-only buyers receive no additional autonomy for that cost. The Cyber 10 Ultra's price is unconfirmed and its arm unvalidated.

Z70 MSRP
$2,599
L50 Ultra MSRP
$1,399
Price delta (arm vs no-arm)
~$1,200
Cyber 10 Ultra price (target)
~$2,100 (unconfirmed USD)
Core L3 autonomy difference
Not documented in reviews

Per-product rollup

How each platform’s claims hold up

Every claim, resolved per product against documented evidence.

ClaimRoborock Saros Z70Dreame Cyber 10 UltraDreame L50 Ultra
Autonomous clutter pickup before cleaningConstrained pickupDemo onlyNo arm
Arm handles real-room object varietyShape-limitedNot real-room testedNo arm
Reduces pre-run floor preparationPrep reducedUnverifiedManual prep required
Arm performance validated beyond demo conditionsMixed reviewsPre-releaseCleaning proven
Premium price reflects arm capability delivered$1,200 arm deltaPrice unconfirmedLevel III at $1,399

Common questions

What readers ask about this comparison.

Q.
Does the Roborock Saros Z70 arm work in a normal home?
Owner reports and press reviews from 12+ months of shipping show the arm succeeds with socks, small toys, and charging cables under 200 g placed openly on mapped floors. WIRED’s review notes it ‘doesn’t quite work’; The New York Times describes it as ‘tries and mostly succeeds.’ The gap captures the current state: real capability with real constraints.
Q.
What objects will the Saros Z70 arm fail to pick up?
Manufacturer documentation recommends pre-clearing phones, glasses, and ceramics. Owner reports add irregular shapes, items flush to walls or chair legs, and objects heavier than roughly 200 g. Fixed grip force with no object identification means it cannot distinguish breakable from toy.
Q.
Is the Dreame Cyber 10 Ultra arm meaningfully better than the Saros Z70?
Manufacturer specs list 500 g lift capacity versus Saros Z70’s ~200 g, and a four-joint arm design. Only documented demo showed ball pickup in a structured booth environment as of June 2026. No independent test of real-room clutter handling exists; whether spec translates to practical performance is undocumented.
Q.
If I do not care about an arm, is the Dreame L50 Ultra reasonable?
The L50 Ultra delivers Level III LiDAR navigation, heated dual-tank mopping, and 15+ months of field-validated reliability at $1,399 MSRP. Core cleaning autonomy documented in press reviews is comparable to arm-equipped models. The arm accounts for the $1,200 price gap.
Q.
Why does the Cyber 10 Ultra score 'diverges' if the arm is more capable?
Verdicts assess claims against documented evidence, not engineering potential. As of June 2026, the Cyber 10 Ultra has not shipped; only documented demo was in a controlled booth. Real-home clutter handling claims cannot be supported or refuted without post-shipping data.
Next up

RTK-plus-camera or LiDAR: two Segway Navimow architectures

Read the comparison
Comparison ID: RV–CMP–7452 · Last reviewed Jun 23, 2026 · Based on owner reports, manufacturer documentation, and firmware release notes