hosted inference · benchmarked & curated

The best model for every specialized task — benchmarked and hosted.

Reranking, grounding, detection, OCR, and the specialist long tail. We benchmark the field for each task and serve the top model — OpenAI-compatible, no GPU to run.

30-second start

# rerank documents against a query — no GPU, no self-hosting
curl https://gigarouter.ai/v1/rerank \
  -H "Authorization: Bearer $GR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"meta-llama/Llama-3.1-8B-Instruct","query":"capital of France",
       "documents":["Paris is the capital of France.","Bananas are yellow."]}'

live catalog

all models →

meta-llama/Llama-3.1-8B-Instruct

Qwen/Qwen3-4B-Instruct-2507

text generation

$0.05 / 1M tokens

Qwen/Qwen3-VL-4B-Instruct

Best for the task

For each specialized task we benchmark the field and serve the top model — not whatever's easiest to host. Need to count objects, detect a species, ground a box, or read a receipt? You get the one that actually wins.

The whole long tail

Rerankers, embedders, grounding, detection, OCR, depth, speech — specialized models with no good hosted API. If it's useful and underserved, we host it. Hundreds and growing.

Drop-in

OpenAI-style endpoints, priced in the unit that fits the task. Point your existing client at the base URL, keep your code.