The LLM Stack Census · Opinion Radar
AI responses, an independent measurement
The LLM Stack Census
Get the results of the next measurement : be notified when the answers change
Alert me
Experimental census · 20 systems · 5 constraints · 6 model families
What software stack do LLMs build by default?
We gave the same architecture briefs to six model families without naming any technology. This report measures their choices—not the objective quality of the architectures.
600 architectures
6 model families
20 systems
99% structured responses
This report is interactive<br>Click any technology, model, or constraint to see the architectures behind each result.
Explore the responses →
The default stack
The choices that appear most often
FrontendReact 83.2%<br>Explore →
Backend languageGo 46.5%<br>Explore →
Backend frameworkFastAPI 19.2%<br>Explore →
Primary databasePostgreSQL 86.2%<br>Explore →
CacheRedis 96.8%<br>Explore →
Queue / event busApache Kafka 31.4%<br>Explore →
SearchElasticsearch 44.0%<br>Explore →
HostingKubernetes 27.9%<br>Explore →
OrchestrationKubernetes 77.4%<br>Explore →
ObservabilityPrometheus 23.0%<br>Explore →
AuthenticationKeycloak 13.3%<br>Explore →
Object storageS3 33.6%<br>Explore →
Model fingerprints
Six models, six architects
ModelAverage componentsFavorite databaseFavorite hostingKubernetesMicroservices<br>Claude (haiku) →14.0PostgreSQLKubernetes71.0%41.0%DeepSeek →14.0PostgreSQLAWS70.7%39.4%Gemini (lite) →7.8PostgreSQLKubernetes78.0%60.0%Le Chat (small) →14.3PostgreSQLEC240.8%21.4%ChatGPT (mini) →15.9PostgreSQLKubernetes72.0%46.0%Grok →7.2PostgreSQLKubernetes48.5%19.2%
Constraint tests
Do models depart from their reflex stack?
ConstraintAverage componentsManaged servicesKubernetesStack changed<br>neutral →12.761.9%66.7%—budget →9.956.8%0.0%85.1%scale →13.763.0%84.9%68.6%eu →12.647.0%83.3%72.5%open_source →12.20.8%83.8%76.0%
Concentration
How many real choices are there in each layer?
FrontendHHI 0.694<br>React 83.2% · 32 distinct choices
Backend languageHHI 0.3<br>Go 46.5% · 25 distinct choices
Backend frameworkHHI 0.086<br>FastAPI 19.2% · 64 distinct choices
Primary databaseHHI 0.744<br>PostgreSQL 86.2% · 31 distinct choices
CacheHHI 0.937<br>Redis 96.8% · 14 distinct choices
Queue / event busHHI 0.166<br>Apache Kafka 31.4% · 60 distinct choices
SearchHHI 0.27<br>Elasticsearch 44.0% · 34 distinct choices
HostingHHI 0.091<br>Kubernetes 27.9% · 147 distinct choices
OrchestrationHHI 0.603<br>Kubernetes 77.4% · 46 distinct choices
ObservabilityHHI 0.087<br>Prometheus 23.0% · 120 distinct choices
AuthenticationHHI 0.044<br>Keycloak 13.3% · 121 distinct choices
Object storageHHI 0.165<br>S3 33.6% · 45 distinct choices
Technology bundles
The most frequent combinations
PostgreSQL + Redis 415Kubernetes + PostgreSQL 345PostgreSQL + React 317Kubernetes + Redis 317React + Redis 285Kubernetes + React 234Go + PostgreSQL 231Go + Kubernetes 219Go + Redis 205GitHub Actions + PostgreSQL 203GitHub Actions + Redis 187Go + React 153
Explore the data
From headline numbers to raw responses
Filter all 600 architectures, compare each variant with its neutral baseline, then expand down to the exact prompt and model response.
Explore all 600 architectures
ModelAll models<br>SystemAll systems<br>ConstraintAll constraints<br>Technology<br>Reset
Expand an architecture to inspect its components and source evidence.
Show 50 more results
JavaScript is required for the interactive explorer.
Methodology
A descriptive census, not a quality ranking
Each brief asks for one production architecture and one decision per role. Variants add one isolated constraint: budget, scale, EU data residency, or self-hostable open source. Raw responses and structured data are downloadable.
Download the JSON data →
Generated on 2026-07-23 07:02