DeepSeek Models on XALEN

Chinese AI lab building efficient reasoning models. Full DeepSeek lineup on XALEN.

Founded: 2023 · HQ: Hangzhou, China · 5 models available

Available Models

DeepSeek R1

Reasoning

State-of-the-art open reasoning model with transparent thinking.

Params: 671B Context: 128K Input: $0.08 Output: $0.15 Latency: ~800ms

DeepSeek R1 0528

Reasoning

Updated R1 with improved accuracy and reduced hallucination.

Params: 671B Context: 128K Input: $0.08 Output: $0.15 Latency: ~800ms

DeepSeek R1 Distill 70B

Reasoning

Distilled reasoning at faster speed and lower cost.

Params: 70B Context: 128K Input: $0.04 Output: $0.08 Latency: ~400ms

DeepSeek V3

Open Source

Efficient MoE model with excellent performance-per-token.

Params: 671B (37B active) Context: 128K Input: $0.05 Output: $0.09 Latency: ~400ms

DeepSeek Coder V2

Code

MoE code model for efficient code generation.

Params: 236B (21B active) Context: 128K Input: $0.03 Output: $0.06 Latency: ~250ms

Quick Start

import XALEN from 'xalen-sdk';

const client = new XALEN({ apiKey: 'xln_test_YOUR_KEY' });

const response = await client.chat.completions.create({
  model: 'deepseek-r1',
  messages: [{ role: 'user', content: 'Hello!' }]
});

Start Using DeepSeek Models

All models available through a single API. No separate accounts needed.

Get API Key

Last updated: 2026-05-21.