3BChat
Llama 3.2 3B Instruct
Meta's small, fast Llama 3.2 in instruct mode. Best for low-latency chat, RAG-style completions, and lightweight tool use. 128K context.
$2.00/hr
Inference Engine
Models
Datasets
Inference Engine
Hosted Inference Endpoints
Click Create Endpoint on any model to get a persistent, OpenAI-API-compatible URL serving that model. Works with cURL, the OpenAI Python SDK, and any OpenAI client unchanged. Billed per hour while the endpoint is running.
1024-dimEmbeddings
BGE-M3 Embeddings
BAAI's multilingual embedding model. Returns 1024-dim vectors usable as drop-in replacements for OpenAI embeddings. Common pick for retrieval (RAG) over African-language corpora.
$1.00/hr
7BChat
Qwen 2.5 7B Instruct
Alibaba's mid-size Qwen 2.5 in instruct mode. Strong reasoning + multilingual performance, including African languages via the broader Qwen pretraining mix. 32K context.
$2.50/hr