Back to list
Product LaunchOctenAI Agent SearchPage Retrieval

Octen Expands Agent Search With Low-Latency Page Retrieval and Broader Image Lookup

VentureBeat reports that Octen launched a search model designed for AI agents and expanded its search and image lookup offerings. Octen reports median server-side retrieval of around 10 milliseconds in lite mode, a figure that applies to the retrieval step rather than to the full request. End-to-end request latency differs, and the supplied report does not give a corresponding figure. The report also distinguishes two aspects of access: the unlimited web-search price applies to a personal plan that uses queuing, while the published model weights carry noncommercial terms. These distinctions matter when reading the announcement: retrieval speed, plan pricing and permitted use of the weights describe different parts of Octen’s offering.

VentureBeat AI

Key Takeaways

  • Octen launched an agent-oriented search model and expanded its search and image lookup offerings, according to VentureBeat.
  • Octen reports median server-side retrieval of around 10 milliseconds in lite mode. The figure does not represent end-to-end request latency.
  • The unlimited web-search price applies to a personal plan with queuing.
  • Octen published model weights under noncommercial terms, a separate consideration from access through the personal search plan.

In-Depth Analysis

What Octen’s 10-Millisecond Figure Measures

The reported figure is specific: Octen says median server-side retrieval takes around 10 milliseconds in lite mode. Each part of that description limits what can be concluded. “Server-side retrieval” identifies the step being measured, while “lite mode” identifies the mode to which the figure applies. “Median” describes the middle of the measured results, not the duration of every retrieval. And because the figure comes from Octen, it should be attributed to the vendor. For readers assessing the model as part of an AI agent, the central distinction is between how long retrieval takes and how long a full request takes. VentureBeat’s account says end-to-end request latency differs, but the supplied report does not give a number for it. It therefore supports a statement about lite-mode server-side retrieval, not a claim that an agent can complete a search request in 10 milliseconds. The expansion of image lookup is part of the same announcement, but the report does not give a separate image lookup timing or say that the 10-millisecond figure applies to it.

What Unlimited Web Search Means for the Personal Plan

The pricing claim has a different scope. VentureBeat reports that the unlimited web-search price is for a personal plan and that the plan uses queuing. Those qualifications are part of the offer as described, not exceptions readers should overlook. “Unlimited” describes the web-search offer, but it does not say how soon a queued request will be served. The report does not specify queue wait times or how queuing affects a full request. That matters when reading the pricing and latency claims together: the existence of a personal plan with unlimited web search does not turn a server-side retrieval measurement into a plan-wide response-time estimate. Nor does the supplied account establish that the unlimited pricing applies to image lookup or use outside the personal plan. Readers can draw a narrower comparison from the announcement: the retrieval metric describes a measured step in a particular mode, while the plan description identifies the offer to which the unlimited web-search price applies and notes that requests are subject to queuing.

What the Published Model Weights Permit

Octen also published model weights, which the report says carry noncommercial terms. Publication and permission are distinct: the weights have been made available, but their release does not establish permission for commercial use. That condition concerns the weights, whereas the personal plan and its queue concern the web-search offer. The supplied account does not set out the full terms for the weights, so it does not support conclusions about individual use cases beyond the reported noncommercial condition. Likewise, the terms on the weights do not by themselves establish the terms of access to Octen’s search service. For readers comparing the two, the first question is which option they are assessing. An evaluation of the web-search offer should account for the personal plan and its queue. An evaluation that depends on using the published weights should start with the noncommercial terms rather than assuming that publication confers commercial rights.

Industry Impact

Octen’s launch gives teams evaluating search for AI agents a reported retrieval figure and an example of why the scope of that figure matters. A comparison with another search offering would need to distinguish server-side retrieval from full request latency. Otherwise, two numbers presented as measures of search speed might describe different parts of a request. The account also does not give enough information to compare the timing of Octen’s web search and image lookup, even though both offerings were expanded.

The announcement raises separate questions about service access and use of the model. For a team considering the web-search plan, the relevant details in this report are its personal scope and queuing. For a team considering the published weights, the relevant detail is the noncommercial condition. These facts do not establish how Octen performs in a production agent workflow or how it compares with competing offerings. They do identify what a fuller evaluation would need to address: the timing of a complete request, the effect of queuing where it applies, and whether the terms cover the intended use.

Frequently Asked Questions

Does Octen complete a search request in about 10 milliseconds?

That is not what the reported figure measures. Octen reports around 10 milliseconds for median server-side retrieval in lite mode. End-to-end request latency differs, and the supplied account does not specify what it is.

Does the unlimited web-search plan use a queue?

Yes. The unlimited web-search price described in the report applies to a personal plan with queuing. The report does not state how long requests spend in that queue.

Are Octen’s published model weights available for commercial use?

The supplied report says the published weights carry noncommercial terms. It does not provide the full terms or identify a separate commercial option for the weights.

Related News

Instagram Rolls Out Content Protection Feature to Help Creators Detect and Review Copycat Videos Across Platform
Product Launch

Instagram Rolls Out Content Protection Feature to Help Creators Detect and Review Copycat Videos Across Platform

Instagram has begun rolling out a new content protection tool aimed at supporting video creators by detecting unauthorized copies of their work. According to reporting from The Verge, the feature functions as an automated detection system designed specifically for eligible creators who enroll in the program. Once activated, the system actively searches for visual matches of original video uploads and notifies creators when matching content is identified across Instagram. This update empowers enrolled creators by giving them the visibility to review duplicated video assets directly within the platform, offering a formal mechanism to track copies of their original multimedia work.

Product Launch

OpenAI Introduces GPT-6.2 Astra and Sol With Better Coding and Lower Costs

OpenAI introduced GPT-6.2 Astra and Sol, describing improvements in professional outputs, coding, computer use, and efficiency at lower cost. The announcement names three offerings where these improvements apply: ChatGPT Work, Codex, and the API. The source presents Astra and Sol together rather than explaining what distinguishes them. It also does not quantify the improvements, identify a comparison baseline, or provide pricing. The announcement therefore establishes the areas OpenAI says have improved and the offerings it names, but it does not show how much better either model performs, how the two compare, or how much users might save on a particular task.

Put Gemini to Work: Free Tasks Features and Enhanced Workspace Integration
Product Launch

Put Gemini to Work: Free Tasks Features and Enhanced Workspace Integration

Google has announced that 'Tasks in Gemini' is now free to use in six countries. This update includes new features launching in the United States, alongside deeper Google Workspace integration designed to help users streamline their workflows and accomplish tasks directly through the Gemini interface.