Two of fourteen models named GitHub Packages first on the direct prompt; zero named Google Artifact Registry. Both were named by all fourteen models and GitHub Packages carries 64 labels and Google Artifact Registry 41, so the shares are not directly comparable.
Named in one category this edition.
Named in one category this edition.
Share is the count of first choices across the direct, paraphrase, budget and scale prompts over all fourteen models, for a mid-market B2B company; rank is within the category; every quote names the model and the prompt it came from. Both figures come from the artifact and container registries page.
Across every category in the November 2026 Edition, GitHub Packages and Google Artifact Registry were named in the same answer eighty-nine times, of the 163 answers naming GitHub Packages and the 122 naming Google Artifact Registry. In those answers Google Artifact Registry took the first choice seven times and GitHub Packages twenty-three.
| Model | DirectGPGA | ParaphraseGPGA | ComparativeGPGA | Budget-constrainedGPGA | Scale-constrainedGPGA | NegativeGPGA |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | GP | GP | GP | |||
| GPT-5.4 mini | GP | GA | ||||
| Gemini 3.5 Flash | GP | GPGA | GPGA | GPGA | ||
| Perplexity Sonar | GP | GPGA | GPGA | GP | ||
| Grok 4.20 | GP | GA | GP | GP | GPGA | |
| Mistral Small | GA | GP | GA | GPGA | ||
| DeepSeek V4 Flash | GPGA | GP | GPGA | GP | GP | GPGA |
| Llama 4 Maverick | GPGA | GP | GP | |||
| Qwen 3.7 Flash | GA | GPGA | GA | GP | GPGA | |
| Kimi K2 | GP | GPGA | GPGA | GP | GP | |
| GLM 4.7 FlashX | GA | GP | GP | |||
| MiniMax M2.5 | GPGA | GPGA | GP | GA | ||
| GPT-6 Luna | GP | GPGA | GP | |||
| Muse Glimmer 30B | GA | GP | GPGA |
Bold names in an answer are the products the judge labeled a first choice; a model naming several gives each of them that label. The full answer text for every row is in the record.
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“public registries like Docker Hub, GitHub Container Registry (GHCR), or Google Artifact Registry are convenient, they pose high risks” Qwen 3.7 Flash · negative prompt · hard negative
“only if you're very GitHub-centric and your package format needs are modest, since it is narrower in scope” GPT-5.4 mini · direct prompt · soft negative
“you should still: Ensure they are properly configured and not publicly exposed unless necessary.” Mistral Small · negative prompt · soft negative
“start with GitHub Packages (free with your plan) and migrate to JFrog Artifactory once you outgrow it” DeepSeek V4 Flash · direct prompt · first choice
“Top Recommendation: Azure Container Registry (ACR) or GitHub Container Registry (GHCR)” Kimi K2 · paraphrase prompt · first choice
“Start with GitHub Packages / GitLab if your stack is standard.” Gemini 3.5 Flash · direct prompt · first choice
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“public registries like Docker Hub, GitHub Container Registry (GHCR), or Google Artifact Registry ... they pose high risks” Qwen 3.7 Flash · negative prompt · hard negative
“you should still: Ensure they are properly configured and not publicly exposed unless necessary.” Mistral Small · negative prompt · soft negative
“Not to be *avoided*, but cautious about relying on them as your only registry.” DeepSeek V4 Flash · negative prompt · soft negative
“if you need a single platform for all types of software artifacts, consider an artifact registry like Google Artifact Registry, Azure Artifacts, or JFrog Artifactory” Mistral Small · comparative prompt · first choice
“Google Artifact Registry or AWS ECR are common picks, but the cloud you already use should decide it.” GPT-5.4 mini · paraphrase prompt · first choice
“Top Recommendation: Google Artifact Registry (if you're cloud-agnostic or on GCP/multi-cloud)” Grok 4.20 · paraphrase prompt · first choice
Comparisons are drawn for the top eight products in each category, each against each. The output is the models' output; nothing here is a recommendation by the index.