ZenMux is excited to welcome Ling-1T, the first flagship non-thinking model from Ant Group's inclusionAI, featuring 1 trillion total parameters with ≈ 50 billion active parameters per token. Built on the Ling 2.0 architecture, Ling-1T is designed to push the limits of efficient reasoning and scalable cognition.
Pre-trained on 20 trillion+ high-quality, reasoning-dense tokens, Ling-1T-base supports up to 128K context length and adopts an evolutionary chain-of-thought (Evo-CoT) process across mid-training and post-training. This curriculum greatly enhances the model's efficiency and reasoning depth, allowing Ling-1T to achieve state-of-the-art performance on multiple complex reasoning benchmarks—balancing accuracy and efficiency.
With ZenMux's Ling-1T API, you can expect:
- Highly Competitive Pricing: Get access to a trillion-parameter model at a fraction of the cost. With input pricing at $0.56 per million tokens and output at $2.24 per million tokens, Ling-1T offers exceptional performance for its price point.
- Massive Context Window: Leverage a 128K context window to process and analyze extensive documents, complex codebases, or lengthy conversations with ease.
- Peak Performance on Demanding Tasks: Ling-1T is optimized for a range of challenging tasks, including: complex code generation, advanced mathematical reasoning, software development and front-end generation.
Our mission has always been to make access to world-class AI simple, stable, and scalable. Through our unified API, developers and enterprises can now seamlessly tap into Ling-1T's core capabilities—without the usual headaches of infrastructure setup or model instability.
Why Ling-1T Matters
Flagship-Level Efficient Reasoning: Ling-1T has been comprehensively evaluated against leading models including DeepSeek-V3.1-Terminus, Kimi-K2-Instruct-0905, GPT-5-main, and Gemini-2.5-Pro. Across code generation, software development, competition-level mathematics, and logical reasoning, Ling-1T consistently demonstrates superior complex reasoning ability. In the AIME 25 benchmark, it extends the Pareto frontier of reasoning accuracy versus reasoning length, showcasing "efficient thinking and precise reasoning."
Visual Reasoning and Front-End Excellence: Ling-1T excels in front-end code generation through a hybrid Syntax–Function–Aesthetics reward mechanism, enabling it to generate not only correct and functional code but also visually aesthetic interfaces. On ArtifactsBench, Ling-1T ranks first among open-source models.
Emergent Intelligence at Trillion-Scale: Scaling to trillion parameters has revealed strong emergent capabilities. In the BFCL V3 tool-use benchmark, Ling-1T achieves approximately 70% tool-call accuracy with only light instruction tuning. The model can interpret complex natural-language instructions, transform abstract logic into functional visual components, generate cross-platform compatible code, and create stylistically controlled multi-lingual content.
Below is a comprehensive summary of Ling-1T's performance across key benchmarks compared to other leading models:
(Note: The benchmark table from the original document should be inserted here)
Why Use Ling-1T on ZenMux
ZenMux is built to make model access simple, stable, and developer-friendly. When you deploy Ling-1T through our platform, you gain access to a suite of features designed to streamline your AI integration and enhance reliability.
Benefits of using Ling-1T via ZenMux:
- Unified Access Experience: Call Ling-1T like any other model without new credentials or SDKs. Our standardized API means you can integrate seamlessly across your existing infrastructure.
- Smart Model Routing: Automatically optimizes latency and throughput, ensuring your requests are handled efficiently and reliably.
- Output Quality Support: Designed to help ensure reliability with ZenMux's upcoming AI model insurance, providing an additional layer of confidence in production deployments.
- Multi-Model Orchestration: Easily combine Ling-1T with other models available on ZenMux, enabling sophisticated workflows that leverage the strengths of multiple AI systems.
In short: you focus on building, ZenMux handles the orchestration.
Get Started Immediately
- Explore: Try Ling-1T via Chat in the ZenMux Studio.
Preview - Integrate: We recommend using the OpenAI SDK or the Anthropic SDK compatibility mode. You can also call the ZenMux API directly. For more details, please check the sample code and API for Ling-1T in the ZenMux Documentation.
from openai import OpenAI
# 1. Initialize the OpenAI client
client = OpenAI(
# 2. Point the base URL to the ZenMux endpoint
base_url="https://zenmux.ai/api/v1",
# 3. Replace with the API Key from your ZenMux User Console
api_key="<your ZENMUX_API_KEY>",
)
# 4. Send a request
completion = client.chat.completions.create(
# 5. Specify the model you want to use in the format "provider/model-name"
model="inclusionai/ling-1t",
messages=[
{
"role": "user",
"content": "What is the meaning of life?"
}
]
)
print(completion.choices[0].message.content)Ready to experience the power of Ling-1T? Start building with our API today and see the difference efficient AI can make.
About ZenMux
ZenMux is the world's first enterprise-grade large model aggregation platform with an insurance payout mechanism. The platform provides one-stop access to the latest models across providers. When issues such as poor output quality or excessive latency occur during use, our intelligent insurance detection and payout mechanism automatically compensates, addressing enterprise concerns around AI hallucinations and unstable quality.