---
type: analysis
slug: eka-s-2-5-picking-claim-needs-a-named-baseline
title: "Eka’s 2.5× picking claim needs a named baseline"
description: "Eka’s reported picking rate is striking, but a multiplier means little without the system it was measured against and the test conditions. The useful next step is a like-for-like trial that reports successful picks per hour, not a headline ratio."
published_at: 2026-10-21T14:00:00.000Z
topics: ["trends"]
source_urls: ["https://techcrunch.com/2026/10/20/eka-is-worth-1-45-billion-thanks-to-a-physical-ai-model-that-has-experts-raving","https://techcrunch.com/2026/10/07/robot-data-startup-mecka-ai-nabs-60m-from-sequoia","https://figure.ai/news/introducing-index"]
author: "Andrey Reshetnikov"
author_profile: https://newruntime.com/owner-profile.md
---
# Eka’s 2.5× picking claim needs a named baseline

Eka’s reported picking rate is striking, but a multiplier means little without the system it was measured against and the test conditions. The useful next step is a like-for-like trial that reports successful picks per hour, not a headline ratio.

## Summary

Eka Robotics has raised $150 million in a Series A at a reported $1.45 billion post-money valuation. The company says its robot picks about 1,500 objects weighing up to one kilogram per hour with more than 95% success. According to dossier finding d1, Eka also describes its rate as roughly 2.5 times a comparison rate, but the available evidence does not identify that baseline or the protocol behind the comparison.

## What happened

The reported figure combines speed and success, two measures that can be difficult to compare across picking systems. A rate depends on the objects, presentation, task definition, handling of failed attempts, and operating conditions; the evidence provided does not specify these details or identify the system behind the 2.5× comparison. Eka’s account describes force-sensitive gripping and a vision-force-action approach, but the claim at issue is not whether those techniques are plausible. It is whether the reported rate beats a relevant alternative under equivalent conditions. Without that context, the multiplier is a company claim, not a usable benchmark.

## What the archive adds

The archive includes reports about Mecka gathering human movement data and Figure Index collecting everyday-action videos for robot training. Those efforts illustrate different parts of the robotics ecosystem, but they do not validate Eka’s throughput claim. The available archive does not provide a directly comparable picking benchmark, so it cannot establish what Eka’s 2.5× figure means against an incumbent system or another manipulation approach.

## Trend signal

Robotics claims increasingly need to connect model capability to measurable task outcomes. Yet throughput numbers only help buyers when the denominator is clear: what task, what objects, what baseline, and what counts as a successful completed pick. A multiplier can make a result sound portable when it may describe a narrow test setup. This makes transparent protocols—not just higher headline rates—important to evaluating progress in physical AI.

## Why now

According to the primary evidence, Eka reports a high picking rate and success figure using a force-sensitive gripper and a vision-force-action model. According to dossier finding d1, the company’s timing reflects a technical bet on combining those elements with simulation and reinforcement learning; this describes its rationale, not independent validation. According to finding d2, investors cite labor constraints and variability in manufacturing, logistics, and food production as reasons flexible automation may be attractive.

## Connections to other events

The central numbers question is the unspecified comparison behind Eka’s reported 2.5× rate. If the baseline used the same objects, task boundaries, operating time, and success criteria, the ratio could indicate a meaningful throughput improvement. If those conditions differ, it cannot tell a buyer whether Eka is faster on their line. The reported 1,500 picks per hour and success rate are useful starting points for a trial, but neither resolves that comparison on its own.

## Why it matters

Buyers should request the raw trial conditions and compare Eka with their current process on the same representative objects. Agree in advance on what counts as a completed pick, include failed attempts and recovery time, and report successful picks per hour over the same operating window. Builders should publish the baseline and protocol alongside any multiplier; doing so makes a result easier to reproduce and harder to misread. Until then, treat the 2.5× figure as a lead for evaluation rather than a procurement conclusion.

## What happens next

- **Eka’s public materials or a customer account will specify the comparison or test conditions behind its reported picking-rate advantage.** (6 months, 45%). Watch for: A company or customer report names the baseline and describes the task and measurement protocol.
- **A buyer evaluating Eka will publish or describe a trial using its own objects and a stated successful-picks-per-hour measure.** (12 months, 40%). Watch for: A named deployment account gives results from a customer-specific comparison with defined success criteria.

## Limits and uncertainty

The evidence pack does not include the full test protocol, identify the comparator behind the 2.5× figure, or provide independent replication. It also does not show how the reported rate and success percentage were measured together. The dossier’s other findings do not supply a comparable picking benchmark, so no conclusion here ranks Eka against specific alternatives. A disclosed, well-matched test could show that the ratio is more informative than the available evidence currently allows.

## Skill to build now

This week, draft a one-page picking benchmark protocol: select representative objects, define a completed pick, fix the test window, count failures and recovery time, and calculate successful picks per hour. Use identical conditions for the candidate system and the process it would replace.

## Reader takeaway

Before relying on Eka’s multiplier, ask for the comparator, object set, trial duration, definition of success, and treatment of recovery time. Then run the same test against your existing process and compare successful picks per hour. That protocol will tell you more about operational value than the ratio alone.

## Open question

Will Eka disclose a baseline and test protocol that make its 2.5× rate comparable to the picking tasks buyers actually need to automate?

## Sources

- [techcrunch.com](https://techcrunch.com/2026/10/20/eka-is-worth-1-45-billion-thanks-to-a-physical-ai-model-that-has-experts-raving)

- [techcrunch.com](https://techcrunch.com/2026/10/07/robot-data-startup-mecka-ai-nabs-60m-from-sequoia)

- [figure.ai](https://figure.ai/news/introducing-index)

- [Eka Robotics Raises $150M to Build Superhuman Intelligence for the Physical World](https://ekarobotics.com/blog/eka-robotics-raises-150m-to-build-superhuman-intelligence-for-the-physical-world)

- [Investing in Eka Robotics](https://a16z.com/announcement/investing-in-eka-robotics/)

- [Picking & Packaging Robots | FANUC America](https://www.fanucamerica.com/applications/picking-packaging)

- [OpenVLA: An Open-Source Vision-Language-Action Model for Robotic Manipulation](https://github.com/openvla/openvla)

- [Eka Robotics](https://ekarobotics.com/)

Author: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Trends](https://newruntime.com/topics/trends.md): Explore the trends topic hub.
- collection: [Analysis](https://newruntime.com/analysis.json): Open the latest authorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)