Tags GA on launch
The full SupraBench capability taxonomy with usage counts. Useful for populating filter dropdowns or mapping your own tags onto ours.
Endpoint
GET
https://api.suprabench.com/v1/tags
No parameters. Returns the entire taxonomy in one shot — typically a few hundred entries, < 30 KB.
Response — array of TagInfo
| Field | Type | Description |
|---|---|---|
| tag | string | Lowercase, hyphenated identifier (e.g. "long-context"). |
| models | integer | Number of models currently carrying this tag (after community vote thresholding). |
| benches | integer | Number of benchmarks tagged with this capability. |
Example
curl https://api.suprabench.com/v1/tags \
-H "Authorization: Bearer $SUPRABENCH_KEY"
tags = requests.get(
"https://api.suprabench.com/v1/tags",
headers={"Authorization": f"Bearer {KEY}"},
).json()
by_popularity = sorted(tags, key=lambda t: -t["models"])
const r = await fetch("https://api.suprabench.com/v1/tags", {
headers: { Authorization: `Bearer ${KEY}` },
});
const tags = await r.json();
const byPopularity = tags.sort((a, b) => b.models - a.models);
req, _ := http.NewRequest("GET",
"https://api.suprabench.com/v1/tags", nil)
req.Header.Set("Authorization", "Bearer "+os.Getenv("SUPRABENCH_KEY"))
resp, _ := http.DefaultClient.Do(req)
defer resp.Body.Close()
var tags []struct {
Tag string `json:"tag"`
Models int `json:"models"`
Benches int `json:"benches"`
}
json.NewDecoder(resp.Body).Decode(&tags)
[
{ "tag": "agentic", "models": 31, "benches": 6 },
{ "tag": "coding", "models": 58, "benches": 12 },
{ "tag": "long-context", "models": 24, "benches": 4 },
{ "tag": "math", "models": 47, "benches": 9 },
{ "tag": "multimodal", "models": 19, "benches": 7 },
{ "tag": "reasoning", "models": 71, "benches": 14 },
{ "tag": "safety", "models": 14, "benches": 5 }
]
About the taxonomy
- Community-driven. Tags appear when users propose them and persist when they accumulate net upvotes. Low-quality tags fall off automatically.
- Lowercase hyphenated.
long-context, neverlongContextor"long context". - Cross-cutting. A tag can apply to both models and benchmarks. The semantics are "this benchmark tests X" or "this model is good at X".
- No hierarchy. Tags are flat.
code-generationandcode-revieware siblings, not children ofcoding. - Cached for 1 hour. The taxonomy doesn't change minute-to-minute.