7*24

Hot topic

Market

Activities

According to 1M AI News monitoring, the Meituan Longcat team has open-sourced LongCat-Next, a native multimodal model based on the MoE architecture with 3 billion activated parameters. It unifies five capabilities—text, visual understanding, image generation, speech understanding, and speech synthesis—within a single autoregressive framework. The model and its accompanying tokenizer are open-sourced under the MIT license, with weights available on HuggingFace. The core design of LongCat-Next is the DiNA (Discrete Native Autoregressive) paradigm: by designing paired tokenizers and decoders for each modality, it converts visual and audio signals into discrete tokens, sharing the same embedding space as text, and completing all tasks through unified next-token prediction. The key component on the visual side, dNaViT (discrete native Resolution Vision Transformer), extracts image features into 'visual words,' supporting dynamic tokenization and decoding. It maintains strong image generation quality even with a 28x compression ratio, showing particular excellence in text rendering. In comparisons with models of similar activated parameter scale (A3B), LongCat-Next's main benchmark performance is as follows: 1. Visual Understanding: MMMU-Pro 60.3 (Qwen3-Omni 57.0, GPT5-minimal 62.7), MathVista 83.1 (Qwen3-Omni 75.9, GPT5-minimal 50.9), MathVision 64.7 (leading all compared models), DocVQA 94.2. 2. Image Generation: GenEval 84.44, LongText-EN 93.15 (FLUX.1-dev 60.70, Emu-3.5 97.60). 3. Programming: SWE-Bench 43.0 (Kimi-Linear-48B 32.8, Qwen3-Next-80B 37.6). 4. Agent Tool Calling: Tau2-Retail 73.68 (Qwen3-Next 57.3), Tau2-Telecom 62.06 (Qwen3-Next 13.2). In a horizontal comparison of unified models for understanding and generation, LongCat-Next's MMMU score of 70.6 leads the second-place NEO-unify (68.9), significantly surpassing previous unified model solutions like BAGEL (55.3) and Ovis-U1 (51.1). The performance on SWE-Bench 43.0 and the Tau2 series of tool-calling benchmarks also indicates that this unified multimodal architecture does not sacrifice pure text and Agent capabilities.

[Meituan Open-Sources LongCat-Next: A 3B Parameter Unified Model for Vision, Generation, and Speech]According to 1M AI News monitoring, the Meituan Longcat team has open-sourced LongCat-Next, a native multimodal model based on the MoE architecture with 3 billion activated parameters. It unifies five capabilities—text, visual understanding, image generation, speech understanding, and speech synthesis—within a single autoregressive framework. The model and its accompanying tokenizer are open-sourced under the MIT license, with weights available on HuggingFace. The core design of LongCat-Next is the DiNA (Discrete Native Autoregressive) paradigm: by designing paired tokenizers and decoders for each modality, it converts visual and audio signals into discrete tokens, sharing the same embedding space as text, and completing all tasks through unified next-token prediction. The key component on the visual side, dNaViT (discrete native Resolution Vision Transformer), extracts image features into 'visual words,' supporting dynamic tokenization and decoding. It maintains strong image generation quality even with a 28x compression ratio, showing particular excellence in text rendering. In comparisons with models of similar activated parameter scale (A3B), LongCat-Next's main benchmark performance is as follows: 1. Visual Understanding: MMMU-Pro 60.3 (Qwen3-Omni 57.0, GPT5-minimal 62.7), MathVista 83.1 (Qwen3-Omni 75.9, GPT5-minimal 50.9), MathVision 64.7 (leading all compared models), DocVQA 94.2. 2. Image Generation: GenEval 84.44, LongText-EN 93.15 (FLUX.1-dev 60.70, Emu-3.5 97.60). 3. Programming: SWE-Bench 43.0 (Kimi-Linear-48B 32.8, Qwen3-Next-80B 37.6). 4. Agent Tool Calling: Tau2-Retail 73.68 (Qwen3-Next 57.3), Tau2-Telecom 62.06 (Qwen3-Next 13.2). In a horizontal comparison of unified models for understanding and generation, LongCat-Next's MMMU score of 70.6 leads the second-place NEO-unify (68.9), significantly surpassing previous unified model solutions like BAGEL (55.3) and Ovis-U1 (51.1). The performance on SWE-Bench 43.0 and the Tau2 series of tool-calling benchmarks also indicates that this unified multimodal architecture does not sacrifice pure text and Agent capabilities.

Meituan Open-Sources LongCat-Next: A 3B Parameter Unified Model for Vision, Generation, and Speech

On March 26th, Solana-based liquidity protocol Meteora announced that its chart-based DLMM (Dynamic Liquidity Management Model) range setting feature is now live on Meteora. This feature allows users to identify support levels, discover resistance levels, and precisely view liquidity positions on charts.

[Meteora Launches Chart-Based DLMM Range Setting Feature]On March 26th, Solana-based liquidity protocol Meteora announced that its chart-based DLMM (Dynamic Liquidity Management Model) range setting feature is now live on Meteora. This feature allows users to identify support levels, discover resistance levels, and precisely view liquidity positions on charts.

Meteora Launches Chart-Based DLMM Range Setting Feature

On March 26, market news reported that the Solana Foundation announced new requirements for the validator delegation program, set to take effect on May 1, 2026. The updates focus on fairness in transaction ordering, prohibiting censorship, implementing stricter temporal rules, and limiting ASN and data center concentration. This adjustment aims to enhance network performance and user experience, with the monitoring platform Ghost assisting in tracking and publishing network health metrics.

[Solana Foundation Updates Validator Delegation Program Requirements, Effective May 1]On March 26, market news reported that the Solana Foundation announced new requirements for the validator delegation program, set to take effect on May 1, 2026. The updates focus on fairness in transaction ordering, prohibiting censorship, implementing stricter temporal rules, and limiting ASN and data center concentration. This adjustment aims to enhance network performance and user experience, with the monitoring platform Ghost assisting in tracking and publishing network health metrics.

Solana Foundation Updates Validator Delegation Program Requirements, Effective May 1

On March 26, Japan's two-year government bond yield climbed to its highest level since 1996, driven by market expectations that the Bank of Japan will raise interest rates in the near future. The yield on Japanese two-year government bonds, which is highly sensitive to monetary policy expectations, rose 1 basis point on Thursday to 1.315%, surpassing the previous high of 1.31% reached last month. The yield on 10-year Japanese government bonds increased by 2 basis points to 2.270%. Markets anticipate that rising oil prices following the outbreak of the conflict in Iran will trigger an inflationary shock. Central banks have warned of persistent price pressures, pushing up short-term yields, while traders have largely dismissed expectations of the Federal Reserve easing policy this year. The rise in oil prices has also put pressure on the Japanese yen, further increasing market expectations that the Bank of Japan may need to continue tightening monetary policy. Overnight index swap data shows that the market is pricing in a 64% chance of the Bank of Japan taking action in April. (JIN10)

[Japan's 2-Year Bond Yield Hits 30-Year High on Rate Hike Expectations]On March 26, Japan's two-year government bond yield climbed to its highest level since 1996, driven by market expectations that the Bank of Japan will raise interest rates in the near future. The yield on Japanese two-year government bonds, which is highly sensitive to monetary policy expectations, rose 1 basis point on Thursday to 1.315%, surpassing the previous high of 1.31% reached last month. The yield on 10-year Japanese government bonds increased by 2 basis points to 2.270%. Markets anticipate that rising oil prices following the outbreak of the conflict in Iran will trigger an inflationary shock. Central banks have warned of persistent price pressures, pushing up short-term yields, while traders have largely dismissed expectations of the Federal Reserve easing policy this year. The rise in oil prices has also put pressure on the Japanese yen, further increasing market expectations that the Bank of Japan may need to continue tightening monetary policy. Overnight index swap data shows that the market is pricing in a 64% chance of the Bank of Japan taking action in April. (JIN10)

Japan's 2-Year Bond Yield Hits 30-Year High on Rate Hike Expectations

On March 26, data disclosed by Kobeissi Letter revealed that in January, countries' holdings of US Treasury debt surged by $34.8 billion, reaching $9.3 trillion, a record second-highest level. Japan, as the largest foreign holder, purchased $39.8 billion worth of US Treasury debt, bringing its total purchases to $1.2 trillion, the highest level since July 2022. The United Kingdom is the second-largest global holder of US debt, increasing its holdings by $29.3 billion to $895.3 billion, a record third-highest level. The United Kingdom is the third-largest global holder of US debt, increasing its holdings by $10.9 billion to $694.4 billion, the highest level since September of last year. Meanwhile, multiple EU countries collectively increased their holdings by $8 billion, reaching $2.1 trillion.

[Foreign Holdings of US Debt Hit Record High]On March 26, data disclosed by Kobeissi Letter revealed that in January, countries' holdings of US Treasury debt surged by $34.8 billion, reaching $9.3 trillion, a record second-highest level. Japan, as the largest foreign holder, purchased $39.8 billion worth of US Treasury debt, bringing its total purchases to $1.2 trillion, the highest level since July 2022. The United Kingdom is the second-largest global holder of US debt, increasing its holdings by $29.3 billion to $895.3 billion, a record third-highest level. The United Kingdom is the third-largest global holder of US debt, increasing its holdings by $10.9 billion to $694.4 billion, the highest level since September of last year. Meanwhile, multiple EU countries collectively increased their holdings by $8 billion, reaching $2.1 trillion.

Foreign Holdings of US Debt Hit Record High

On March 26th, Bo Shen, founder of BitTorrent Capital, announced on social media that he is publicly seeking clues related to his personal wallet theft and has established a bounty program.

Bo Shen stated that his personal wallet was hacked in 2022, resulting in a loss of approximately $42 million. Over the past three years since the incident, his team has been continuously tracking the case and has now obtained more key clues and evidence. The on-chain flow path of the stolen assets has also become gradually clearer. This bounty program is open to all individuals and institutions, with no restrictions on identity, background, or format. Those who make substantial contributions to the recovery of the assets will receive a reward of 10%–20% of the finally recovered amount, with the specific proportion determined by the extent of their actual contribution.

[BitTorrent Capital Founder Offers Bounty for Stolen Wallet Clues]On March 26th, Bo Shen, founder of BitTorrent Capital, announced on social media that he is publicly seeking clues related to his personal wallet theft and has established a bounty program.

Bo Shen stated that his personal wallet was hacked in 2022, resulting in a loss of approximately $42 million. Over the past three years since the incident, his team has been continuously tracking the case and has now obtained more key clues and evidence. The on-chain flow path of the stolen assets has also become gradually clearer. This bounty program is open to all individuals and institutions, with no restrictions on identity, background, or format. Those who make substantial contributions to the recovery of the assets will receive a reward of 10%–20% of the finally recovered amount, with the specific proportion determined by the extent of their actual contribution.

BitTorrent Capital Founder Offers Bounty for Stolen Wallet Clues

On March 25th, according to researcher Vadym's analysis, DeFi is projected to generate approximately $8 billion in on-chain revenue by 2025. The largest source will be AMM trading fees, estimated at $4.2 billion, with Uniswap, Meteora, and Raydium accounting for 62% of this. Lending interest will be the second largest contributor, at around $1.76 billion, with money markets like Aave and Morpho contributing over 60% of DeFi's total TVL. However, approximately half of the lending demand is attributed to cyclical leverage operations.

Real World Assets (RWA) are expected to contribute $600 million to $900 million, with US Treasuries making up about 41% of the RWA market. Perpetual contract funding rates are projected to contribute around $300 million, primarily from Ethena. Notably, over half of stablecoin deposits within the Ethereum ecosystem are yielding less than US Treasury rates. Potential revenue streams such as insurance underwriting and on-chain options remain underdeveloped. The analysis uses Sky (formerly MakerDAO) as an example, highlighting that approximately 70% of its revenue originates from off-chain assets, indicating that TradFi yields are flowing into DeFi through permissioned channels at an accelerated pace.

[DeFi Projected to Generate $8 Billion in On-Chain Revenue by 2025, Over Half of Stablecoin Deposits Yield Less Than US Treasuries]On March 25th, according to researcher Vadym's analysis, DeFi is projected to generate approximately $8 billion in on-chain revenue by 2025. The largest source will be AMM trading fees, estimated at $4.2 billion, with Uniswap, Meteora, and Raydium accounting for 62% of this. Lending interest will be the second largest contributor, at around $1.76 billion, with money markets like Aave and Morpho contributing over 60% of DeFi's total TVL. However, approximately half of the lending demand is attributed to cyclical leverage operations.

Real World Assets (RWA) are expected to contribute $600 million to $900 million, with US Treasuries making up about 41% of the RWA market. Perpetual contract funding rates are projected to contribute around $300 million, primarily from Ethena. Notably, over half of stablecoin deposits within the Ethereum ecosystem are yielding less than US Treasury rates. Potential revenue streams such as insurance underwriting and on-chain options remain underdeveloped. The analysis uses Sky (formerly MakerDAO) as an example, highlighting that approximately 70% of its revenue originates from off-chain assets, indicating that TradFi yields are flowing into DeFi through permissioned channels at an accelerated pace.

DeFi Projected to Generate $8 Billion in On-Chain Revenue by 2025, Over Half of Stablecoin Deposits Yield Less Than US Treasuries

On March 26th, U.S. Commodity Futures Trading Commission (CFTC) Chair Michael Selig stated that driving the return of genuine cryptocurrency perpetual futures business to the United States is a core component of the CFTC's innovation policy agenda.

[CFTC Chair: Return of Crypto Perpetual Futures Business to US is Core to CFTC's Innovation Agenda]On March 26th, U.S. Commodity Futures Trading Commission (CFTC) Chair Michael Selig stated that driving the return of genuine cryptocurrency perpetual futures business to the United States is a core component of the CFTC's innovation policy agenda.

CFTC Chair: Return of Crypto Perpetual Futures Business to US is Core to CFTC's Innovation Agenda

On March 26th, U.S. Commodity Futures Trading Commission (CFTC) Chair Michael Selig stated that driving the genuine return of cryptocurrency perpetual futures business to the United States is a core component of the CFTC's innovation policy agenda.

[CFTC Chair: Bringing Crypto Perpetual Futures Business Back to the US is Core to CFTC's Innovation Agenda]On March 26th, U.S. Commodity Futures Trading Commission (CFTC) Chair Michael Selig stated that driving the genuine return of cryptocurrency perpetual futures business to the United States is a core component of the CFTC's innovation policy agenda.

CFTC Chair: Bringing Crypto Perpetual Futures Business Back to the US is Core to CFTC's Innovation Agenda

The Sandbox, World Tourism Forum Institute and TourismX are collaborating to develop Tourismverse, a virtual neighborhood which hosted on The Sandbox platform. It aims to bring all countries to visit the destinations virtually and interact tourism content through different activities.

Detail

【The Sandbox to partner with World Tourism Forum Institute and TourismX to launch Tourismverse】The Sandbox, World Tourism Forum Institute and TourismX are collaborating to develop Tourismverse, a virtual neighborhood which hosted on The Sandbox platform. It aims to bring all countries to visit the destinations virtually and interact tourism content through different activities.

The Sandbox to partner with World Tourism Forum Institute and TourismX to launch Tourismverse

Article

US Bitcoin and Ethereum ETFs See Net Outflows

Bhutan Transfers 519 Bitcoin to New Addresses

OpenAI Completes Pre-training for Next-Gen Model Spud; Altman Steps Down from Safety Oversight to Focus on Data Centers

LiteLLM Poisoning: Karpathy Calls 'One pip install' Incident 'Most Terrifying in Software'

BTC broke through 72000  USDT

The Sandbox to partner with World Tourism Forum Institute and TourismX to launch Tourismverse

Descargo de responsabilidad：El contenido anterior no representa las posiciones de HTX.，HTX no ofrece ninguna recomendación de trading.。

Artículos relacionados

Todos los comentarios0Lo más recientePopular

Temas populares

Artículos relacionados