10 local LLMs, one brand prompt, 8 setups: what I learned generating product copy on an ASUS GX10

I needed German product descriptions for a B2B workwear and corporate fashion shop (Shopware 6, 10,000+ articles). Requirements: correct German, no invented properties, and the client's brand rules. Everything runs locally on one ASUS GX10 (NVIDIA GB10, 128 GB memory shared by CPU and GPU, no discrete graphics card).

I ran the test in three steps. Every text was checked against the supplier data and scored out of 10 against a fixed rubric, with Claude (Anthropic) as the reviewer and a written reason per text. That's a reasoned judgement, not a measurement.

Full report with all tables, per-article scores and runtime settings: GX10 Product Copy Test

Ten local models, 30 articles, same data, same prompt. The Gemma 4 family won (Gemma 4 26B: 288/300, with 12B and 31B right behind). Typical failures elsewhere: dropped safety standards on hi-vis jackets, prompt examples copied in as industries, an invented "business customer offer" in meta descriptions.

The three largest models (71 to 86 GB) ranked below the 18 GB Gemma 4 26B.

The client, a designer, wrote her brand voice as a prompt: plain language, banned clichés, strict fact rules. Same 30 articles, top five models:

Scoring was stricter because tone counted, so look at the spread: 7 points with the plain prompt, 111 with the real one. Benchmark with your production prompt.

First I fixed the pipeline (50 new articles from 50 product types, shop-ready HTML plus SEO fields). What helped most:

Then the same pipeline with different builds and runtimes:

Two brand-new models from October 2026 didn't come close: NVIDIA Nemotron 3.5 Bolt scored 305, Aleph Alpha Kolibri-1 290. Repeated runs of the same Gemma setup landed between 416 and 421, so a few points are noise. Gaps of 100 are not.

The serving call, for reference. Thinking off, JSON schema where structured output is needed:

body = { " model ": " gemma4-31b-qat ", " messages ": messages, " temperature ": 0.3, " max_tokens ": 1200, " chat_template_kwargs ": { " enable_thinking ": False }, " response_format ": { " type ": " json_schema ", " json_schema ": { " name ": " answer ", " schema ": schema }}} Enter fullscreen mode Exit fullscreen mode Takeaways Benchmark with your real prompt and your worst data. Pick the build, not just the model. Use a batching server for bulk jobs. Check mandatory facts with fixed rules. Keep a human spot check. Everything else (step-by-step tables, per-article scores for all 11 runs, the vLLM settings for Gemma, Nemotron and Kolibri on a GB10, one article compared across models) is in the full report.

I'm a Shopware freelancer. If you've measured QAT against post-hoc quants on other tasks, I'd like to compare notes in the comments.