Description
Kimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods across various contexts, including short, long, and reinforcement learning (RL) scaling regimes. At its core is Kimi Delta Attention (KDA)—a refined version of Gated DeltaNet that introduces a more efficient gating mechanism to optimize the use of finite-state RNN memory. Kimi Linear achieves superior performance and hardware efficiency, especially for long-context tasks. It reduces the need for large KV caches by up to 75% and boosts decoding throughput by up to 6x for contexts as long as 1M tokens.
API Usage Examples
OpenAI Compatible Endpoint
Use this endpoint with any OpenAI-compatible library. Model: MoonshotAI: Kimi Linear 48B A3B Instruct (moonshotai/kimi-linear-48b-a3b-instruct)
curl https://api.ridvay.com/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer YOUR_API_KEY" -d '{
"model": "moonshotai/kimi-linear-48b-a3b-instruct",
"messages": [
{
"role": "user",
"content": "Explain the capabilities of the MoonshotAI: Kimi Linear 48B A3B Instruct model"
}
],
"temperature": 0.7,
"max_tokens": 1024
}'Supported Modalities
- Text
API Pricing
- Input: 0.5$ / 1M tokens
- Output: 0.6$ / 1M tokens
Token Limits
- Max Output: 1,048,576 tokens
- Max Context: 1,048,576 tokens
Subscription Tiers
- free
- pro
- ultimate