Reported by The Robot Report on . Summarized and analyzed by XAIPART Editorial.
What happened
TwelveLabs Inc. released Pegasus 1.6, the first version of its Pegasus model built specifically to understand egocentric video, which is shot from the point of view of the person performing a task.S1 The company describes Pegasus 1.6 as an expansion of its video-understanding platform into physical AI, aimed at helping robotics and autonomous-system teams convert raw footage into training data.S1
Co-founder and CEO Jae Lee stated that Pegasus 1.6 can segment actions in time, identify left- and right-hand behavior, describe how the surrounding environment changes, and provide a coarse read of limb motion.S1 Lee said the model is not intended to replace a robot's tactile or actuator-level sensing, but to be combined with a team's own sensor data to build trajectories and learning policies.S1
Pegasus 1.6 supports five workflows: action segmentation and labeling, dense caption labeling, quality scoring, search and curation, and consent and compliance flagging.S1 It also extends Pegasus 1.5's Time-Based Metadata (TBM) capability, improves entity recognition for hands, objects, and tools across clips, and adds faster, more cost-efficient processing for high-volume video.S1 The model works with existing video archives and accepts still images in addition to video, and TwelveLabs stated it does not require a specific camera or proprietary hardware.S1
Lee said the company is collaborating with robotics labs working on dexterity and manipulation tasks such as packaging, assembly, cleaning, and semiconductor quality control, and that egocentric video is more scalable to gather than teleoperation data.S1
| Metric | Reported value | Source |
|---|---|---|
| Model release | Pegasus 1.6 | [S1] |
| Company founded | 2021 | [S1] |
| Headquarters | Seoul-based | [S1] |
| Target video type | Egocentric (operator point of view) | [S1] |
| Supported workflows | 5 (action segmentation, dense caption labeling, quality scoring, search/curation, consent flagging) | [S1] |
| Hardware requirement | No specific camera or proprietary hardware required | [S1] |
| Input modalities | Video and still images | [S1] |
| Pricing, model size, latency benchmarks | Not disclosed | [S1] |
Background timeline
- - TwelveLabs founded; later built Marengo and Pegasus as a full-stack video intelligence platform. [S1]
- - Pegasus 1.5 released, introducing Time-Based Metadata (TBM) functionality and attracting new enterprise customers. [S1]
- - TwelveLabs announces Pegasus 1.6, the first model built for egocentric video and physical AI workflows. [S1]
XAIPART take
Technical impact
For design engineers building data pipelines for robot learning, Pegasus 1.6 is positioned as a pre-processing layer rather than a perception stack.XAIPART analysis The model targets the labeling bottleneck: time-stamped action labels, hand-object interaction descriptions, and quality scoring are exactly the annotations that otherwise require human reviewers at scale.XAIPART analysis Because the source notes no required camera hardware and adds still-image support, integration with existing IMU, force/torque, and encoder feedback is presented as additive, not replacing on-robot sensing.XAIPART analysis
Market impact
For procurement and design-in decisions, the relevant question is whether vendor-managed video-understanding APIs can replace in-house annotation teams.XAIPART analysis The source does not disclose pricing, throughput, or latency, so total cost of ownership versus a human-label workflow cannot be compared from this brief.XAIPART analysis Mention of packaging, assembly, cleaning, and semiconductor QC suggests TwelveLabs is targeting industrial OEMs that already maintain large egocentric footage libraries, which is the buyer segment most likely to evaluate a paid tier.XAIPART analysis
Action
Engineering leads building imitation-learning or vision-language-action (VLA) datasets should benchmark Pegasus 1.6 on a representative slice of their own egocentric footage before committing to a pipeline redesign.XAIPART analysis Procurement teams should request per-minute or per-hour pricing, data-retention terms, and consent-flagging audit logs, since the source leaves all three undisclosed.XAIPART analysis For now, treat the model as a labeling and curation utility layered on top of existing camera, IMU, and tactile sensor stacks rather than as a sensing replacement.XAIPART analysis
XAIPART Forecast
XAIPART expects that, by Q1 2027, TwelveLabs will publish either per-minute processing pricing for Pegasus 1.6 or a published latency benchmark on egocentric video clips of at least 10 minutes; if neither appears by 2027-03-31, this forecast is wrong.XAIPART analysis
Facts are as reported by the cited source. The design-in, sourcing, and forward-looking statements above are XAIPART's analysis, not the source's claims. Uncertainty: pricing, model size, and latency benchmarks are not disclosed; customer names and integration details beyond packaging, assembly, cleaning, and semiconductor QC are not disclosed.
Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.
Alternatives et cycle de vie
| Référence | Relation | Cycle de vie | Fondement |
|---|---|---|---|
| — | Aucune alternative référencée DONNÉES REQUISES | — | Nécessite match_type avant que toute mention de prédécesseur ne puisse apparaître |
Les mentions « prédécesseur » et « alternative » proviennent uniquement d'un type de correspondance enregistré, d'un statut de cycle de vie ou du texte source. Lorsqu'aucune correspondance n'est enregistrée, aucune mention n'est affichée. Nos propres annonces ne constituent pas une référence de marché.
À surveiller
| Affirmation | Seuil | Par | Statut |
|---|---|---|---|
| XAIPART expects that, by Q1 2027, TwelveLabs will publish either per-minute processing pricing for Pegasus 1.6 or a published latency benchmark on egocentric video clips of at least 10 minutes | neither appears by 2027-03-31 | Pending |
Les prévisions qui se révèlent erronées restent visibles et signalées. Nous ne les supprimons pas.
Paysage concurrentiel
- 2 annonces actives dans cette catégorie sur XAIPART
- 2 fabricants représentés
- 68 fabricants mentionnés dans nos archives d'actualités au cours des 30 derniers jours
Les chiffres reflètent uniquement le catalogue XAIPART — ils ne représentent pas des parts de marché.
Références concernées
Aucune correspondance dans le catalogue. Cette information ne correspond à aucune référence actuellement listée sur XAIPART. Nous ne déduisons pas de correspondances et n'affichons pas de liste fictive.
Pression sur les coûts par dimension
| Dimension | FRED series ID | Unit | Lag | As of |
|---|---|---|---|---|
| Input cost | PCOPPUSDM, PALUMUSDM, PNICKUSDM | USD/mt | 52 d | |
| Supply/demand | PCU334413334413, IPG3344S | index(1982=100) / index(2017=100) | 46 d | |
| FX & energy | DEXCHUS, DCOILBRENTEU, DHHNGSP | rate / USD/bbl / USD/MMBtu | 5 d |
No free authoritative interface for: gold · silver · rare earth · polysilicon · lithium carbonate · market share — listed rather than approximated.
Comment ce contenu a été produit
Traçabilité du processus — récupération, contrôle de la source, modèle et contrôle
| Source fetched | The Robot Report · www.therobotreport.com · 12,376 chars · fetch status success |
|---|---|
| Source sanity (S1) | ok — passed before any model call (no s1_ block recorded) |
| Model / prompt | MiniMax-M3 · prompt Blog Agent - News Summary (v2.8-final) · 9,191 tokens · 14.9s · run completed |
| Gate | verdict hold · 6/6 applicable checks passed · spec v2.9 · C10 ratio 0.011581 · C10 rounds 0 |
| Time contract | 4 <time datetime> values in this brief — 0 violations (+08:00 or date-only) |
Chaque ligne ci-dessus est produite par le pipeline, et non rédigée à la main.
Commentaires
Soyez le premier à commenter.
🔐 Veuillez vous connecter pour publier un commentaire.
Vous n'avez pas de compte ? vous inscrire