Superpower Daily

Superpower Daily

Your reading space

Sign in
Perplexity Says It Trusts GPT-6 Astra With Production SystemsCognition Releases SWE-2 for Devin, Claiming Lower-Cost Coding...Pentagon Says It Has Moved 90% of Classified AI Work Off Anthro...Google Completes Mechanize Talent Deal, Bringing Staff Into Mid...Harvey Raises $550 Million Series H at $15.5 Billion ValuationGottheimer and Lawler Introduce Bill for Verifiable AI-Agent Se...Perplexity Says It Trusts GPT-6 Astra With Production SystemsCognition Releases SWE-2 for Devin, Claiming Lower-Cost Coding...Pentagon Says It Has Moved 90% of Classified AI Work Off Anthro...Google Completes Mechanize Talent Deal, Bringing Staff Into Mid...Harvey Raises $550 Million Series H at $15.5 Billion ValuationGottheimer and Lawler Introduce Bill for Verifiable AI-Agent Se...
Swipe

Databricks Adds DeepSeek V4.1 Flash for Text-and-Image AI Requests

The new platform option gives organizations another way to use DeepSeek’s newly released multimodal model, though Databricks says individual workspaces may receive staged releases later.

By 2 min read
Databricks Adds DeepSeek V4.1 Flash for Text-and-Image AI Requests
Databricks Adds DeepSeek V4.1 Flash for Text-and-Image AI Requests

Listen to this story

The audio brief

About 1:28
0:001:28
Read transcript
Databricks has added DeepSeek V4.1 Flash to Unity Gateway, giving customers a new way to send text-and-image requests through its Foundation Model APIs. The model appeared on Databricks on September 11, 2026, one day after DeepSeek announced it and made it available through its own API. The practical change is consolidation. Teams already working inside Databricks can access DeepSeek’s new multimodal model through the platform interface, rather than setting up a separate route just for visual inputs. DeepSeek also offers direct access under the model name deepseek-flash, with what it describes as native visual understanding. The underlying system is large: DeepSeek describes V4.1 Flash as a 552-billion-parameter mixture-of-experts model. That does not mean all of those parameters are active on every request. DeepSeek says 8 billion input parameters and 16 billion output parameters are active, as part of an architecture aimed at faster inference and higher throughput. Databricks has also associated the release with performance on OfficeQA Pro V2, saying the model reaches that benchmark’s Pareto frontier. But availability is not necessarily immediate for every customer. Databricks stages releases, so some workspaces may receive the model a week or more after the initial release date. Customers also remain responsible for checking compliance with applicable terms. The key constraint, then, is whether a particular workspace has actually received the update when a team needs it.

Story brief

3 key points

Databricks has added DeepSeek V4.1 Flash to Unity Gateway via its Foundation Model APIs, expanding access to a new multimodal model for customers already using its platform. DeepSeek describes the model as a 552-billion-parameter mixture-of-experts system, with 8 billion input and 16 billion output parameters active, and offers direct access as deepseek-flash. The practical benefit is consolidated text-and-image...

  1. 01

    DeepSeek announced V4.1 Flash on September 10; Databricks listed it on Unity Gateway the following day.

  2. 02

    Databricks supports text and image inputs; DeepSeek offers native visual understanding through its direct API.

  3. 03

    The model uses a 552-billion-parameter mixture-of-experts design, with 8 billion input and 16 billion output parameters active.

Organizations using Databricks now have a new route for AI requests that combine text and images. DeepSeek V4.1 Flash became available on Unity Gateway through Databricks’ Foundation Model APIs on September 11, one day after DeepSeek announced the model and opened access through its own API.

Databricks lists V4.1 Flash as supporting text and image inputs on Unity Gateway. DeepSeek says the model has native visual understanding and is available on its API under the model name deepseek-flash.

DeepSeek describes V4.1 Flash as a 552-billion-parameter mixture-of-experts model. It says 8 billion input parameters and 16 billion output parameters are active, and presents the design as part of a new architecture family built for faster inference and higher throughput.

What the two providers currently offer

  • Databricks: V4.1 Flash on Unity Gateway with text and image inputs through Foundation Model APIs.
  • DeepSeek: direct API access using the deepseek-flash model name, with native multimodal support.

The September 11 release note does not mean every Databricks workspace receives the model simultaneously. Databricks says its releases are staged and an account might not be updated until a week or more after an initial release date. It also says customers are responsible for ensuring compliance with applicable terms.

For teams already using Databricks’ model interface, the addition places DeepSeek’s new multimodal model alongside their existing platform access. The immediate constraint is rollout timing: a model can be listed in release notes before a particular workspace receives the update.

Sources

  1. deepseek.comIntroducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
  2. docs.databricks.comDeepSeek 4.1 Flash is on Databricks the day it ships. On OfficeQA Pro V2, it lands on the Pareto frontier, making it a strong open-source option for enterprise document reasoning.

Loading discussion...

Related coverage

Continue reading