DeepSeek-V4 Now on SiliconFlow: Million-Token Context Intelligence

TL; DR: DeepSeek-V4 series is now available on SiliconFlow. This release introduces two powerful Mixture-of-Experts models with groundbreaking 1M-token context windows and hybrid attention architecture that reduces inference costs by 73%. DeepSeek-V4-Pro achieves a Codeforces rating of 3206 and 93.5% on LiveCodeBench, establishing itself as the best open-source model available today. Start building with SiliconFlow's API today to explore the Million-Token Context Intelligence.

Overview: Million-Token Context Intelligence

DeepSeek-V4 brings two strong Mixture-of-Experts (MoE) language models: DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated), and both supporting a context length of one million tokens.

As a next-generation MoE model family, DeepSeek-V4 sets a new benchmark for unprecedented long-context efficiency, advanced reasoning capabilities, and state-of-the-art performance across coding, mathematics, and agentic tasks, enabling developers to achieve breakthrough results in complex AI applications with greater efficiency and reliability.

Try DeepSeek-V4-Pro & DeepSeek-V4-Flash on SiliconFlow

• Cost-effective Pricing: DeepSeek-V4-Pro: $0.145 / $1.74 / $3.48 per 1M tokens; DeepSeek-V4-Flash: $0.028 / $0.14 / $0.28 per 1M tokens.

• Seamless Integration: Instant compatibility with your existing development ecosystem: deploy via SiliconFlow's OpenAI-Compatible API through Cline, Gen-CLI, Kilo Code, Roo Code ; Anthropic-Compatible API with Claude Code; plug into agents like OpenClaw, Hermes Agent; ready-to-use in Dify, Janitor AI, Chub AI, ChatHub, Chatbox, Sider; and also available through OpenRouter.

Key Features & Innovations

DeepSeek-V4 brings revolutionary hybrid attention architecture and architectural innovations, achieving both massive context windows and dramatically reduced inference costs.

• Two Model Variants: DeepSeek-V4-Pro (1.6T parameters, 49B activated) and DeepSeek-V4-Flash (284B parameters, 13B activated) for different performance-efficiency trade-offs.

• Three Reasoning Modes: Non-think for fast responses, Think High for complex problem-solving, and Think Max for pushing reasoning boundaries.

• Hybrid Attention Architecture: Combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA), requiring only 27% of inference FLOPs and 10% of KV cache compared to DeepSeek-V3.2.

• Manifold-Constrained Hyper-Connections (mHC): Strengthens residual connections to enhance signal propagation stability across layers while preserving model expressivity.

• Muon Optimizer: Enables faster convergence and greater training stability during the pre-training phase.

• Trained on 32T: Comprehensive pre-training on diverse, high-quality data followed by domain-specific expert cultivation and unified model consolidation.

DeepSeek-V4-Pro-Max on Benchmarks

• LiveCodeBench: 93.5% pass rate, outperforming leading closed-source models such as Opus 4.6 & Gemini 3.1 Pro.

• Codeforces Rating: 3206, highest among all frontier models, establishing new state-of-the-art for open-source models.

• SWE-Verified: 80.6%, matches Claude Opus 4.6 Max.

• MMLU-Pro: 87.5% accuracy, demonstrating strong knowledge capabilities across diverse domains.

DeepSeek-V4-Pro-Max consistently outperforms previous open-source models and bridges the gap with leading closed-source models in reasoning, coding, mathematics, and agentic tasks.

Real-World Applications

• Advanced Code Development: DeepSeek-V4 excels at code generation, debugging, and complex algorithmic problem-solving across multiple programming languages.

• Long-Document Analysis: The 1M-token context window enables comprehensive analysis of entire codebases, legal documents, research papers, and technical documentation without truncation.

• Agentic AI Systems: Ideal for building autonomous agents that can plan, reason, and execute complex multi-step tasks.

Get Started Immediately

1. Explore: Try DeepSeek-V4-Pro and DeepSeek-V4-Flash in the SiliconFlow playground.

2. Integrate: Use our OpenAI-compatible API. Explore the full API specifications in the SiliconFlow API documentation.

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "deepseek-ai/DeepSeek-V4-Pro",
    "messages": [
        {
            "role": "user",
            "content": "an island near sea, with seagulls, moon shining over the sea, light house, boats int he background, fish flying over the sea"
        }
    ],
    "stream": True,
    "max_tokens": 4096,
    "enable_thinking": False,
    "thinking_budget": 4096,
    "min_p": 0,
    "stop": "1",
    "temperature": 0.7,
    "top_p": 0.7,
    "top_k": 50,
    "frequency_penalty": 0.5,
    "n": 1,
    "response_format": { "type": "json_object" },
    "tools": [
        {
            "type": "function",
            "function": {
                "name": "<string>",
                "description": "<string>",
                "parameters": {},
                "strict": False
            }
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "deepseek-ai/DeepSeek-V4-Pro",
    "messages": [
        {
            "role": "user",
            "content": "an island near sea, with seagulls, moon shining over the sea, light house, boats int he background, fish flying over the sea"
        }
    ],
    "stream": True,
    "max_tokens": 4096,
    "enable_thinking": False,
    "thinking_budget": 4096,
    "min_p": 0,
    "stop": "1",
    "temperature": 0.7,
    "top_p": 0.7,
    "top_k": 50,
    "frequency_penalty": 0.5,
    "n": 1,
    "response_format": { "type": "json_object" },
    "tools": [
        {
            "type": "function",
            "function": {
                "name": "<string>",
                "description": "<string>",
                "parameters": {},
                "strict": False
            }
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "deepseek-ai/DeepSeek-V4-Pro",
    "messages": [
        {
            "role": "user",
            "content": "an island near sea, with seagulls, moon shining over the sea, light house, boats int he background, fish flying over the sea"
        }
    ],
    "stream": True,
    "max_tokens": 4096,
    "enable_thinking": False,
    "thinking_budget": 4096,
    "min_p": 0,
    "stop": "1",
    "temperature": 0.7,
    "top_p": 0.7,
    "top_k": 50,
    "frequency_penalty": 0.5,
    "n": 1,
    "response_format": { "type": "json_object" },
    "tools": [
        {
            "type": "function",
            "function": {
                "name": "<string>",
                "description": "<string>",
                "parameters": {},
                "strict": False
            }
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

Business or Sales Inquiries →

Join our Discord community now →

Follow us on X for the latest updates →

Explore all available models on SiliconFlow →