AGP Picks
View all

Feedoptimise moves AI feed optimization onto self-hosted GPUs

Oct. 28, 2026
Feedoptimise moves AI feed optimization onto self-hosted GPUs
By AI, Created 12:02 UTC, Oct 28, 2026, AGP -

Feedoptimise is shifting its free AI models from a cloud provider to GPUs it manages in its own data center, starting today for all customers. The change adds a larger free monthly AI quota, image-based prompts and full prompt control while keeping product-feed data inside Feedoptimise’s infrastructure.

Why it matters: - Feedoptimise is cutting reliance on third-party AI infrastructure, which gives the platform more predictable capacity and costs. - Retailers get more included AI usage without separate API keys or AI billing accounts. - The move matters for feed quality because product content can be generated and adjusted directly inside the feed management workflow.

What happened: - Feedoptimise moved its free AI models from a cloud provider to GPUs in a data center it manages. - The platform now runs Alibaba’s open-weight Qwen 3.8 model on its own hardware. - The change is available to all Feedoptimise customers starting today. - Feedoptimise says it is the first product feed management platform to run AI models on GPU hardware it operates itself.

The details: - Every plan now includes free AI requests equal to twice the plan’s monthly product limit. - A 10,000-product plan includes 20,000 AI requests per month. - Feedoptimise caches results and reprocesses a product only when source data or settings change, so unchanged products do not use requests. - Users who need more can raise their monthly limit for an additional fee in the months when they use the extra requests. - The new Feedoptimise-AI Prompt modifier lets users write the full prompt and reference any attribute in the feed. - Common uses include rewriting titles to a set pattern, writing channel-specific descriptions, mapping product types to a channel taxonomy, filling missing attributes such as material or gender, scoring title quality and translating content for new markets. - The Feedoptimise-AI Image Prompt modifier sends one or more product images to the model with the prompt and returns a single value, a sentence or structured JSON. - Retailers can use image prompts to distinguish lifestyle shots from model shots and to extract color and material from product photos. - Qwen 3.8 supports image and text input. - Feedoptimise’s self-hosted models work with only the product data retailers already send in their feeds. - Feedoptimise processes that data on servers it operates and does not share it with an external AI provider or use it to train AI models. - Users can also connect commercial models such as ChatGPT, Gemini, Claude, Grok or DeepSeek with their own API key. - Users can connect more than 400 models through OpenRouter and compare results with Feedoptimise’s built-in A/B testing. - Existing users of the AI Title Optimiser, AI Descriptions Creator and AI Attributes Extractor keep their current setup. - Those three modifiers now run on Qwen 3.8 starting today. - The same GPUs also run BiRefNet, the open-source image segmentation model behind the Feedoptimise Image Editor. - Feedoptimise plans to add more open-weight models to its self-hosted lineup. - Feedoptimise added its first self-hosted open model, Qwen 2.5, in January 2025.

Between the lines: - Moving from rented GPU capacity to owned hardware gives Feedoptimise more control over when it upgrades models and how it allocates compute. - The larger included AI quota lowers the barrier for retailers that want to automate more feed tasks without adding separate AI spend. - Self-hosting may appeal to retailers and agencies that want a privacy-first setup for product data. - The emphasis on open-weight models suggests Feedoptimise is trying to reduce dependence on commercial AI vendors while keeping optional access to them.

What’s next: - Feedoptimise plans to add more open-weight models to its self-hosted stack. - Retailers can try the new tools on a free trial. - The company will continue supporting both self-hosted and commercial model workflows through the platform.

The bottom line: - Feedoptimise is turning AI feed optimization into a core, included part of its platform, with self-hosted GPUs, image prompting and a bigger free monthly quota for every plan.

Disclaimer: This article was produced by AGP Wire with the assistance of artificial intelligence based on original source content and has been refined to improve clarity, structure, and readability. This content is provided on an “as is” basis. While care has been taken in its preparation, it may contain inaccuracies or omissions, and readers should consult the original source and independently verify key information where appropriate. This content is for informational purposes only and does not constitute legal, financial, investment, or other professional advice.

Sign up for:

Books, Poets, and Cats

The daily local news briefing you can trust. Every day. Subscribe now.

By signing up, you agree to our Terms & Conditions.

Share this page:

Advanced Search Options

Search for:

Search scope:

Type:

Search in:

Date range:

The last

Sort by:

Sign up for:

Books, Poets, and Cats

The daily local news briefing you can trust. Every day. Subscribe now.

By signing up, you agree to our Terms & Conditions.