Three of fourteen models named HAProxy first on the direct prompt; two named AWS Application Load Balancer. HAProxy was named by fourteen of the fourteen models and AWS Application Load Balancer by ten and HAProxy carries 61 labels and AWS Application Load Balancer 14, so the shares are not directly comparable.
Named in three categories 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 load balancers page.
Across every category in the November 2026 Edition, HAProxy and AWS Application Load Balancer were named in the same answer forty-three times, of the 191 answers naming HAProxy and the 51 naming AWS Application Load Balancer. In those answers AWS Application Load Balancer took the first choice five times and HAProxy twelve.
| Model | Direct | Paraphrase | Comparative | Budget-constrained | Scale-constrained | Negative |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | AA | |||||
| GPT-5.4 mini | AA | |||||
| Gemini 3.5 Flash | ||||||
| Perplexity Sonar | AA | |||||
| Grok 4.20 | ||||||
| Mistral Small | ||||||
| DeepSeek V4 Flash | ||||||
| Llama 4 Maverick | ||||||
| Qwen 3.7 Flash | ||||||
| Kimi K2 | ||||||
| GLM 4.7 FlashX | ||||||
| MiniMax M2.5 | AA | |||||
| GPT-6 Luna | ||||||
| Muse Glimmer 30B | AA |
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.
“Older or unpatched self-hosted solutions (e.g., legacy HAProxy, Apache, or custom LVS setups): These can suffer from single points of failure” Grok 4.20 · negative prompt · soft negative
“HAProxy\u2019s official docs label some releases EOL; don\u2019t assume an old install is still receiving fixes” GPT-6 Luna · negative prompt · soft negative
“Running a single NGINX or HAProxy instance as your load balancer creates a single point of failure” Kimi K2 · negative prompt · soft negative
“Consistently the fastest performer in benchmarks with lowest latency and memory footprint; battle-tested in production for decades.” DeepSeek V4 Flash · comparative prompt · first choice
“The best load balancer for a company with a limited budget is HAProxy, as it is a cost-efficient, open-source deployment option.” Llama 4 Maverick · budget prompt · first choice
“HAProxy is the best overall load balancing software for most teams because its open-source core is free and production-grade” Claude Haiku 4.5 · comparative prompt · first choice
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Five of six in this category shown.
“AWS ALB is unsuitable for workloads requiring L4 forwarding, ultra-low-latency pass-through, or non-HTTP protocols; use NLB or the appropriate L4 service instead.” Perplexity Sonar · negative prompt · soft negative
“ALB is not the right choice when the workload is non-HTTP (use NLB)” Muse Glimmer 30B · negative prompt · soft negative
“I'd recommend starting with AWS ALB or Azure Load Balancer (depending on your cloud provider)” Claude Haiku 4.5 · direct prompt · first choice
“AWS-centric company: AWS Application Load Balancer for web apps” GPT-5.4 mini · direct prompt · first choice
“Cloud-first, mostly HTTP(S) APIs, limited ops staff | AWS ALB, Azure Application Gateway, Google Cloud Load Balancing” MiniMax M2.5 · direct prompt · alternative
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.