# Model Related Articles

HTX News Center provides the latest articles and in-depth analysis on "Model", covering market trends, project updates, tech developments, and regulatory policies in the crypto industry.

Rhythm X Zhihu Event Guests Announced, Comprehensive Coverage of the New Financial Model Brought by AI Agent from Academic, Institutional, and Individual Perspectives

In a notable incident, GPT creator Sam Altman’s home was targeted in two separate attacks—one involving a Molotov cocktail and the other gunfire—though no one was injured. The timing is significant: it occurred just after OpenAI’s latest model release, following the debut of Anthropic’s powerful model Mythos, and shortly after Hermes 3 outperformed OpenClaw on multiple benchmarks. These events underscore a broader trend: the AI industry is accelerating at a pace that now feels overwhelming. Model update cycles have shrunk from quarters to weeks, sometimes days. What was considered state-of-the-art just a month ago may now be surpassed by multiple competitors. Hermes overtook OpenClaw not in some distant future—but just last week. This acceleration is more than a trend; it’s becoming a structural force. When the speed of technological progress exceeds society’s ability to process it, frustration and fear can manifest in extreme ways. The attacks on Altman’s home serve as a stark signal: uncontrolled technological change can lead to uncontrolled emotional responses.

marsbit2 days ago 07:28

Rhythm X Zhihu Event Guests Announced, Comprehensive Coverage of the New Financial Model Brought by AI Agent from Academic, Institutional, and Individual Perspectives

marsbit2 days ago 07:28

Tsinghua's Prediction 2 Years Ago Is Becoming Global Consensus: Meta and Two Other Major AI Institutions Have Reached the Same Conclusion

Summary: In a remarkable validation of Chinese AI research, Meta and METR have independently reached conclusions that align perfectly with the "Density Law" proposed by a Tsinghua University and FaceWall Intelligent team two years ago. Published in Nature Machine Intelligence in late 2025, the law states that the computational power required to achieve a specific level of AI performance halves every 3.5 months. This convergence was starkly evident in April 2026. METR reported that AI capabilities are doubling every 88.6 days, while Meta's new model, Muse Spark, demonstrated it could match the performance of a model from the previous year using less than one-tenth of the training compute. When plotted, the growth curves from all three sources—using different metrics (parameters, compute, task length)—show an almost identical exponential slope. The findings have profound implications: AI inference costs are collapsing faster than anticipated, powerful edge-computing AI is becoming rapidly feasible, and the industry's strategy of simply scaling model size is becoming economically inefficient. The Chinese team, which has been building its "MiniCPM" model series based on this law since 2024, is seen as having a significant two-year lead in practical engineering experience, marking a rare instance where Chinese researchers pioneered a fundamental predictive trend in AI.

marsbit04/13 12:14

Tsinghua's Prediction 2 Years Ago Is Becoming Global Consensus: Meta and Two Other Major AI Institutions Have Reached the Same Conclusion

marsbit04/13 12:14

Claiming the "Happy Horse": Alibaba's AI Lays Out the "Eight Trigrams Formation"

Alibaba has officially claimed the "HappyHorse" (HappyHorse-1.0) AI video generation model, which recently topped the global benchmark on Artificial Analysis with an Elo score of 1357. Developed by Alibaba’s ATH (Alibaba Token Hub) innovation unit, the model is notable for its ability to generate high-definition video with synchronized audio and sound effects from text input, significantly improving motion coherence and reducing production time and cost. This launch is part of a broader acceleration in Alibaba’s AI strategy. In late March and early April, the company released three flagship models in quick succession: Qwen3.5-Omni, Wan2.7-Image, and Qwen3.6-Plus. The latter broke global daily call volume records with 1.4 trillion tokens processed shortly after release. Alibaba has also undergone significant organizational restructuring to support its AI ambitions. In March, it established the ATH business group, led by CEO Wu Yongming, to integrate AI development, cloud services, and application deployment. Further changes in April included forming a group-level technology committee and consolidating the Tongyi Lab into a dedicated AI model division. The company is investing heavily in AI, with plans to spend over 380 billion RMB on cloud and AI infrastructure over three years. Its self-developed GPUs have already seen mass production. While the market has responded positively to these moves, challenges remain in balancing centralized control with operational flexibility and maintaining team stability amid rapid changes.

marsbit04/11 04:07

Claiming the "Happy Horse": Alibaba's AI Lays Out the "Eight Trigrams Formation"

marsbit04/11 04:07

Wang Tao's First Large Model Brings Meta Back to the Table

Meta has unveiled Muse Spark, the first model from its new Muse series, marking a significant step in its AI strategy under the leadership of Alexandr (Wang) since his appointment ten months ago. Designed to be compact and fast, Spark is a natively multimodal reasoning model that integrates visual information into its core architecture, enabling advanced capabilities like visual chain-of-thought reasoning. It excels in areas such as medical reasoning, visual and chart understanding, and scientific reasoning, though it lags in coding and long-form tasks compared to rivals like GPT-5.4 and Gemini 3.1 Pro. The model will power Meta AI across platforms like WhatsApp, Instagram, and Facebook, with private API access for select partners. The release follows Meta’s shift from its earlier Llama series and significant internal investment, including a restructured AI stack and expanded infrastructure. Spark’s introduction spurred a 6.5% rise in Meta’s stock, reflecting market optimism. Meta plans to develop larger models in the Muse series, potentially adopting a hybrid open/closed-source approach moving forward.

marsbit04/09 10:57

Wang Tao's First Large Model Brings Meta Back to the Table

marsbit04/09 10:57

Meeting at the Pinnacle of Generalist: 30 Billion in 30 Days, What Did Qianxun AI Do Right?

Qianxun Intelligence, a Chinese embodied AI and robotics startup, completed two major funding rounds totaling 3 billion RMB within 30 days in early 2026, backed by prominent investors including Shunwei Capital (Lei Jun) and Yunfeng Capital (Jack Ma). Founded in January 2024 by a team with expertise in robotics, AI, and commercialization, the company focuses on developing general-purpose embodied AI models. Its open-source model, Spirit v1.5, surpassed competitors in performance benchmarks, demonstrating strong zero-shot generalization capabilities for complex tasks. The company follows a scaling law approach similar to large language models (LLMs), leveraging massive diverse datasets—including internet videos, wearable device data, and teleoperation data—to train its Vision-Language-Action (VLA) model. Qianxun employs a multi-source data engine, collecting over 200,000 hours of real-world interaction data, with plans to reach 1 million hours by 2026. It uses low-cost wearable devices for efficient data acquisition and emphasizes real-world deployment for continuous data feedback. The company has deployed robots like "Xiao Mo" in industrial settings (e.g., battery production lines for CATL) and commercial scenarios (e.g., as baristas in JD.com malls), using operational data to refine its models. This "commercialize while iterating" strategy supports both revenue generation and model improvement, positioning Qianxun to compete globally in embodied AI.

marsbit04/07 04:05

Meeting at the Pinnacle of Generalist: 30 Billion in 30 Days, What Did Qianxun AI Do Right?

marsbit04/07 04:05

Performance Surpasses Opus! Anthropic Leaked Document Reveals: The New Generation Super Model Claude Mythos Is Already in Testing

According to leaked internal documents from Anthropic, the company's highly anticipated next-generation AI model, Claude Mythos, is currently in a secret testing phase. The documents reveal a new model tier named "Capybara," which represents a major technological leap, featuring a larger scale and superior intelligence that surpasses the flagship Claude Opus model. The leak also highlights significant concerns within Anthropic regarding unprecedented cybersecurity risks associated with Claude Mythos, prompting a cautious release strategy to balance advanced capabilities with safety. This development is expected to significantly raise the benchmark for large language models, intensifying competition in the AI industry and pushing the evolution of models toward deeper logical reasoning and complex task handling. The official release date for Claude Mythos remains unannounced.

marsbit03/27 03:52

Performance Surpasses Opus! Anthropic Leaked Document Reveals: The New Generation Super Model Claude Mythos Is Already in Testing

marsbit03/27 03:52

"Borrowing a Chicken to Lay Eggs"? Apple Secures Google Gemini License: Crafting the Strongest On-Device AI Through Model "Distillation"

Apple has reportedly secured full access to Google's Gemini model, not for simple API integration, but to perform "model distillation"—a technique that allows Apple to train smaller, more efficient on-device AI models by learning from Gemini’s advanced reasoning processes. This approach enables Apple to develop localized AI systems that operate entirely on-device, improving speed and safeguarding user privacy. The distilled models are expected to power a significantly upgraded Siri in iOS 27, enabling deeper interaction, complex question-answering, document understanding, and practical task execution—all without requiring cloud connectivity. While leveraging Google’s model in the short term, Apple continues to develop its own foundational AI models internally, adopting a dual strategy of near-term collaboration and long-term independence. This move represents Apple’s effort to balance cutting-edge AI performance with its strict privacy and efficiency standards, potentially reshaping the landscape of on-device artificial intelligence.

marsbit03/26 01:08

"Borrowing a Chicken to Lay Eggs"? Apple Secures Google Gemini License: Crafting the Strongest On-Device AI Through Model "Distillation"

marsbit03/26 01:08

How to Earn $300,000 a Year by Counting Musk's Tweets?

An anonymous trader known as "noovd" has reportedly earned over $345,000 in profit since July 2024 by accurately predicting the number of tweets Elon Musk would post each week on the prediction market Polymarket. The user placed 1,281 bets, with the largest single win exceeding $136,000. Noovd’s strategy focused exclusively on betting "Yes" on narrow, often undervalued tweet-count ranges (e.g., 260–279 or 420–439 tweets per week), which were typically priced as low as 2–24 cents due to market underestimation. Using a proprietary quantitative model trained on Musk’s tweet data since 2022, noovd identified mispriced probabilities—often realizing 8–20x expected value advantages. Key to the model’s success was incorporating variables like day-of-week patterns, real-world events (e.g., SpaceX launches, Tesla earnings), and mid-week recalculations to narrow the probable outcome range. The trader’s high-conviction, concentrated bets—often risking thousands per position—allowed for outsized returns, such as a 4,311% gain on a $1,569 bet. Noovd’s edge stemmed from deep data analysis, disciplined execution, and exploiting persistent market inefficiencies. The case illustrates how specialized expertise in a narrow domain can yield significant returns, emphasizing data over intuition in prediction markets.

比推02/21 14:56

How to Earn $300,000 a Year by Counting Musk's Tweets?

比推02/21 14:56

a16z: Behind the Palantir-ization of Everything, a Doomed Imitation Show

In the article "a16z: The Palantir-ization of Everything and the Doomed Imitation Game," Marc Andrusko critiques the trend of AI startups adopting Palantir's model—deploying engineers on-site, offering highly customized solutions, and securing large contracts. While appealing for landing seven-figure deals, this approach often fails to replicate Palantir’s success. Palantir’s uniqueness lies in its integrated platform, elite talent, and focus on mission-critical sectors like defense and intelligence, where high stakes justify deep customization. Most imitators lack these conditions, risking evolution into service-based consultancies with unsustainable economics. Andrusko advises founders to use embedded engineering as scaffolding for product development, prioritize reusable platform primitives, and honestly assess scalability and margins. True Palantir-like success requires a blend of product vision, strategic market selection, and patient capital—not mere imitation.

marsbit01/19 03:18

a16z: Behind the Palantir-ization of Everything, a Doomed Imitation Show

marsbit01/19 03:18

Just 6 Days After Launching ChatGPT Health, OpenAI Is Surpassed on Its Own Medical Benchmark

In a significant development in the AI healthcare sector, Baichuan Intelligence has surpassed OpenAI's GPT-5.2 High on the HealthBench benchmark—a medical evaluation dataset created by OpenAI with input from 260+ doctors across 60 countries—just six days after OpenAI launched ChatGPT Health. Baichuan's new model, Baichuan-M3, achieved a top score of 65.1 and also led in the more challenging HealthBench Hard subset, while demonstrating the lowest hallucination rate (3.5%) without relying on external tools. Key to M3’s performance is its Fact Aware RL technique, which improves diagnostic accuracy by balancing factual precision with proactive questioning. The model avoids both over-confident errors and overly vague responses. Additionally, Baichuan introduced SCAN-bench, a new evaluation framework designed to simulate real doctor-patient interactions. In tests, M3 outperformed human specialists in areas like safety stratification, clarity, and diagnostic questioning, partly due to its ability to integrate knowledge across medical disciplines. Baichuan is now rolling out the model via its consumer product Baixiaoying (百小应), offering tailored interfaces for both doctors and patients. The company emphasizes a focus on "serious medicine," prioritizing complex areas like oncology over general wellness, aiming to augment—not just assist—medical professionals. According to CEO Wang Xiaochuan, enhancing AI’s capability in high-stakes medical scenarios is crucial for building user trust and advancing toward AGI through deeper biological understanding.

marsbit01/14 02:31

Just 6 Days After Launching ChatGPT Health, OpenAI Is Surpassed on Its Own Medical Benchmark

marsbit01/14 02:31

活动图片

Bitcoin

Project Updates