BU, a leading AI research institute, has recently introduced two impressive models, Ernie 4.5 and Ernie X1, that are challenging the dominance of the big players in the AI industry.

Ernie 4.5 is a natively multimodal model that outperforms GPT-4.5 at just 1% of the cost. It exhibits comprehensive improvements in understanding, generation, reasoning, and memory, along with notable enhancements in hallucination prevention, logical reasoning, and coding abilities.

Ernie X1, BU's first reasoning AI model, is matching the performance of DeepSeek's R1 at half the price. This cost-efficient model showcases impressive performance in various benchmarks, including problem-solving, logical deduction, and geometric reasoning.

The pricing of these models is highly competitive, with Ernie 4.5 costing $55 for 1 million input tokens and $2.20 for 1 million output tokens, significantly cheaper than GPT-4.5. Similarly, Ernie X1 is priced at 50% of the cost of DeepSeek R1, with 28 cents for 1 million input tokens and $110 for 1 million output tokens.

Furthermore, BU has announced that Ernie X1 will be made open-source later this year, allowing broader access and adoption of this powerful reasoning model.

The performance of these models has been thoroughly tested across various benchmarks, including coding, math, logical reasoning, and problem-solving. The results demonstrate the impressive capabilities of Ernie 4.5 and Ernie X1, showcasing their potential to drive innovation and impact in the AI industry.

The pricing of Ernie 4.5 and Ernie X1 models is highly cost-efficient compared to their state-of-the-art counterparts. Ernie 4.5, a multimodal model, is priced at $0.55 for 1 million input tokens and $2.20 for 1 million output tokens, which is drastically cheaper than the pricing structure of GPT-4.5. Similarly, the Ernie X1 reasoning model is priced at $0.28 for 1 million input tokens and $110 for 1 million output tokens, which is 50% of the price of the DeepSeek R1 model.

This cost-efficient pricing makes these models highly accessible and enables broader adoption and economic benefits. The CEO of BU, Robert Lee, has emphasized that cost reduction and continuous innovation are crucial for AI scalability and real-world impact. By improving efficiency and automation, the AI models themselves can drive broader adoption and economic benefit, which is why BU's decision to release these models and make them open-source later this year is highly commendable.

Ernie 4.5 is a natively multimodal model that outperforms GPT-4.5 at just 1% of the cost. This impressive model exhibits comprehensive improvements in understanding, generation, reasoning, and memory, along with notable enhancements in hallucination prevention, logical reasoning, and even coding abilities.

Ernie 4.5 achieves strong performance across multiple modalities, showcasing its versatility and capabilities. The model demonstrates significant advancements in areas such as:

• Understanding: Ernie 4.5 has enhanced its ability to comprehend and process information from various sources, including text, images, and other modalities.
• Generation: The model has improved its generation capabilities, allowing it to produce high-quality and coherent content across a wide range of tasks.
• Reasoning: Ernie 4.5 has strengthened its logical reasoning skills, enabling it to tackle complex problems and make informed decisions.
• Memory: The model has enhanced its memory capabilities, allowing it to maintain context and retain relevant information throughout interactions.

Additionally, Ernie 4.5 has made notable improvements in hallucination prevention, ensuring the reliability and trustworthiness of its outputs. The model's logical reasoning and coding abilities have also been enhanced, making it a valuable tool for a variety of applications.

The cost-efficiency of Ernie 4.5 is another significant advantage, as it is priced at $0.55 for 1 million input tokens and $2.20 for 1 million output tokens, which is drastically cheaper than the pricing structure of GPT-4.5. This cost-effectiveness, combined with its impressive performance, makes Ernie 4.5 a highly attractive option for organizations and individuals seeking advanced AI capabilities.

Ernie X1 is BU's first reasoning AI model that is matching DeepSeek's R1 performance at half the price. This model showcases impressive reasoning capabilities, making it a cost-efficient alternative to state-of-the-art models.

The Ernie X1 is priced at 28 cents for 1 million input tokens and $110 for 1 million output tokens, which is 50% of the price of DeepSeek R1. Despite the lower cost, the model demonstrates strong performance in solving complex word problems, geometric equations, and logical deduction tasks.

In the provided tests, the Ernie X1 model was able to accurately solve a train problem, determine the length of a dividing line in a triangular field, and identify the truth tellers and liars in a logical reasoning scenario. The model's step-by-step reasoning process showcased its ability to apply relevant mathematical concepts and logical analysis to arrive at the correct solutions.

The impressive performance and cost-efficiency of the Ernie X1 model make it a compelling option for organizations and individuals seeking powerful reasoning capabilities at a more accessible price point. BU's commitment to making this model open-source later this year further enhances its accessibility and potential for broader adoption in the AI industry.

We will be assessing the performance of Baidu's Ernie 4.5 and Ernie X1 models across various benchmarks. Ernie 4.5 is a multimodal model that outperforms GPT-4.5 at a fraction of the cost, while Ernie X1 is a reasoning-focused model that matches the performance of DeepSeeK R1 at half the price.

Ernie 4.5 Benchmarks

1. Responsive Web App: Ernie 4.5 was able to generate a functional personal finance tracker app, including code snippets for the user interface, data storage, and visualization.

2. Game of Life: The model successfully implemented the Game of Life, a grid-based simulation, demonstrating its capabilities in grid-based programming and efficient state updates.

3. SVG Butterfly: Ernie 4.5 generated an accurate SVG representation of a butterfly with symmetrical wings and simple styling, showcasing its vector programming and design skills.

4. Logical Error Debugging: The model identified and fixed the logical errors in a provided Python function, demonstrating its code debugging and problem-solving abilities.

Ernie X1 Benchmarks

1. Train Problem: Ernie X1 solved a word problem involving the meeting time and location of two trains, showcasing its strong reasoning and problem-solving skills.

2. Geometric Equation: The model successfully solved a geometric problem involving the division of a triangular field, using the appropriate formulas and theorems.

3. Integer Solution: Ernie X1 demonstrated its ability to set up and analyze equations to find integer solutions, even when no valid combination was found.

4. Logical Deduction: The model used logical reasoning and consistency analysis to correctly identify the truth-teller and the liar among the given scenarios.

Overall, both Ernie 4.5 and Ernie X1 have demonstrated impressive performance across a wide range of benchmarks, showcasing their capabilities in areas such as coding, reasoning, problem-solving, and logical deduction. The cost-efficiency of these models, along with Baidu's plan to make Ernie X1 open-source, makes them compelling options in the AI landscape.

Both the Ernie 4.5 and Ernie X1 models from BU have demonstrated impressive capabilities across a range of tasks, from coding and math to logical reasoning and problem-solving. The Ernie 4.5 model, a natively multimodal model, has outperformed the more expensive GPT-4.5 at just 1% of the cost. The Ernie X1, BU's first reasoning AI model, has matched the performance of the DeepSeek R1 at half the price.

These models showcase BU's commitment to cost-efficiency and continuous innovation, which are crucial for driving broader AI adoption and real-world impact, as emphasized by BU's CEO, Robert Lee. The fact that BU will be making the Ernie X1 model open-source later this year is an exciting development, as it will allow more people to access and utilize this powerful reasoning model.

Overall, the performance and cost-effectiveness of these BU models are truly impressive, and they are certainly worth keeping an eye on as the AI industry continues to evolve.