7*24

Hot topic

Market

Activities

According to monitoring by 1M AI News, AI inference infrastructure company Fireworks AI has released a preview of Fireworks Training, expanding from a pure inference platform to an integrated platform for training and deployment. Fireworks AI was founded by Lin Qiao, a former Meta engineer involved in building PyTorch, and is currently valued at $4 billion, processing 15 trillion tokens daily. The platform offers three tiers: 1. Training Agent: Designed for product teams without ML infrastructure, allowing them to describe tasks and upload data to complete the entire process from training to deployment, currently supporting only LoRA; 2. Managed Training: Aimed at ML engineers, supporting SFT, DPO, and reinforcement learning fine-tuning, including full parameter training; 3. Training API: Targeted at research teams, allowing customization of loss functions and training loops, supporting algorithms such as GRPO and DAPO with full parameter training scales ranging from single-node Qwen3 8B to Kimi K2.5 (trillion parameters) on 64 NVIDIA B200s. Fireworks AI's production inference clients, AI programming tools Cursor, Vercel, and Genspark, have completed cutting-edge reinforcement learning training on this platform. Vercel trained an automatic error correction model for its code generation product v0, achieving a 93% error-free code generation rate, compared to only 62% with Sonnet 3.5, and improved end-to-end latency by 40 times compared to the previously used closed-source model. Genspark fine-tuned the trillion-parameter open-source model Kimi K2 with reinforcement learning to build a deep research agent, increasing tool usage by 33% and reducing costs by 50%. Cursor completed distributed reinforcement learning training for Composer 2 across 3 to 4 clusters globally (currently ranked first on CursorBench), sharing the same GPU pool for training and production inference. Fireworks AI emphasizes its core technological differentiation in numerical consistency between training and inference. MoE (Mixture of Experts) models are numerically more fragile than dense models, where minor changes in hidden states can flip expert routing and amplify cascading effects. Fireworks has published the KL divergence values between training and inference for all supported models, all below 0.01.

[Fireworks AI Launches Preview of Training Platform Supporting Trillion-Parameter Full Training]According to monitoring by 1M AI News, AI inference infrastructure company Fireworks AI has released a preview of Fireworks Training, expanding from a pure inference platform to an integrated platform for training and deployment. Fireworks AI was founded by Lin Qiao, a former Meta engineer involved in building PyTorch, and is currently valued at $4 billion, processing 15 trillion tokens daily. The platform offers three tiers: 1. Training Agent: Designed for product teams without ML infrastructure, allowing them to describe tasks and upload data to complete the entire process from training to deployment, currently supporting only LoRA; 2. Managed Training: Aimed at ML engineers, supporting SFT, DPO, and reinforcement learning fine-tuning, including full parameter training; 3. Training API: Targeted at research teams, allowing customization of loss functions and training loops, supporting algorithms such as GRPO and DAPO with full parameter training scales ranging from single-node Qwen3 8B to Kimi K2.5 (trillion parameters) on 64 NVIDIA B200s. Fireworks AI's production inference clients, AI programming tools Cursor, Vercel, and Genspark, have completed cutting-edge reinforcement learning training on this platform. Vercel trained an automatic error correction model for its code generation product v0, achieving a 93% error-free code generation rate, compared to only 62% with Sonnet 3.5, and improved end-to-end latency by 40 times compared to the previously used closed-source model. Genspark fine-tuned the trillion-parameter open-source model Kimi K2 with reinforcement learning to build a deep research agent, increasing tool usage by 33% and reducing costs by 50%. Cursor completed distributed reinforcement learning training for Composer 2 across 3 to 4 clusters globally (currently ranked first on CursorBench), sharing the same GPU pool for training and production inference. Fireworks AI emphasizes its core technological differentiation in numerical consistency between training and inference. MoE (Mixture of Experts) models are numerically more fragile than dense models, where minor changes in hidden states can flip expert routing and amplify cascading effects. Fireworks has published the KL divergence values between training and inference for all supported models, all below 0.01.

Fireworks AI Launches Preview of Training Platform Supporting Trillion-Parameter Full Training

On April 7, according to monitoring by Onchain Lens, Bit Digital has just staked 43,335 ETH, valued at approximately $91.34 million.

[Bit Digital Stakes 43,335 ETH Worth Approximately $91.34 Million]On April 7, according to monitoring by Onchain Lens, Bit Digital has just staked 43,335 ETH, valued at approximately $91.34 million.

Bit Digital Stakes 43,335 ETH Worth Approximately $91.34 Million

On April 7, CoinDesk reported that Paul Atkins, the Chairman of the U.S. Securities and Exchange Commission (SEC), stated on Monday that the agency is set to propose a 'crypto regulation' initiative to further clarify its regulatory approach to the cryptocurrency industry, delineating which transactions may be considered securities and which may not. Chairman Paul Atkins mentioned that the new 'crypto regulation' has been submitted to the White House Office of Information and Regulatory Affairs, indicating that it is just one step away from formal release. He noted that the rule-making is primarily based on the Securities Act of 1933 and will address issues related to financing and startup exemptions. After the Q&A session, he also indicated that the SEC plans to soon launch the long-awaited 'innovation exemption.'

[SEC Chair Announces Upcoming Regulatory Rules on Cryptocurrency Financing]On April 7, CoinDesk reported that Paul Atkins, the Chairman of the U.S. Securities and Exchange Commission (SEC), stated on Monday that the agency is set to propose a 'crypto regulation' initiative to further clarify its regulatory approach to the cryptocurrency industry, delineating which transactions may be considered securities and which may not. Chairman Paul Atkins mentioned that the new 'crypto regulation' has been submitted to the White House Office of Information and Regulatory Affairs, indicating that it is just one step away from formal release. He noted that the rule-making is primarily based on the Securities Act of 1933 and will address issues related to financing and startup exemptions. After the Q&A session, he also indicated that the SEC plans to soon launch the long-awaited 'innovation exemption.'

SEC Chair Announces Upcoming Regulatory Rules on Cryptocurrency Financing

On April 7, traders involved in related shipping reported that since the outbreak of hostilities weeks ago, Iran has not allowed any liquefied natural gas (LNG) carriers to pass through the Hormuz Strait, a ban that could exacerbate global supply shortages. On Monday, two tankers loaded with Qatari LNG appeared to be preparing to leave the Persian Gulf, but turned back hours later. Traders stated that these two vessels were not granted permission by Iranian officials. This sudden turnaround highlights that since the U.S. and Israel began airstrikes against Iran in late February, the Hormuz Strait has evidently been in a state of blockade. The traffic of vessels through the strait has significantly decreased, although tankers and other ships still occasionally pass through, usually requiring Iranian permission—while about one-fifth of global LNG supply remains cut off. Sources indicate that currently, fully loaded LNG carriers are scattered throughout the Persian Gulf, either because they have not yet engaged in negotiations with Iran regarding passage or have not received approval. Vessel tracking data shows that more than a dozen fully loaded LNG carriers are in a state of detention in the region. It is also possible that some vessels have evaded monitoring by turning off their transponders or that signal interference has led to inaccurate tracking. However, both traders and vessel tracking data indicate that only one LNG carrier has managed to exit the Hormuz Strait earlier this month—but it was not loaded with cargo.

[Iran Blocks Hormuz Strait for Weeks, No LNG Ships Allowed to Depart]On April 7, traders involved in related shipping reported that since the outbreak of hostilities weeks ago, Iran has not allowed any liquefied natural gas (LNG) carriers to pass through the Hormuz Strait, a ban that could exacerbate global supply shortages. On Monday, two tankers loaded with Qatari LNG appeared to be preparing to leave the Persian Gulf, but turned back hours later. Traders stated that these two vessels were not granted permission by Iranian officials. This sudden turnaround highlights that since the U.S. and Israel began airstrikes against Iran in late February, the Hormuz Strait has evidently been in a state of blockade. The traffic of vessels through the strait has significantly decreased, although tankers and other ships still occasionally pass through, usually requiring Iranian permission—while about one-fifth of global LNG supply remains cut off. Sources indicate that currently, fully loaded LNG carriers are scattered throughout the Persian Gulf, either because they have not yet engaged in negotiations with Iran regarding passage or have not received approval. Vessel tracking data shows that more than a dozen fully loaded LNG carriers are in a state of detention in the region. It is also possible that some vessels have evaded monitoring by turning off their transponders or that signal interference has led to inaccurate tracking. However, both traders and vessel tracking data indicate that only one LNG carrier has managed to exit the Hormuz Strait earlier this month—but it was not loaded with cargo.

Iran Blocks Hormuz Strait for Weeks, No LNG Ships Allowed to Depart

On April 7, according to monitoring data from Farside, the US Bitcoin spot ETF had a net inflow of $471.4 million yesterday, while the Ethereum ETF saw a net inflow of $120.2 million.

[US Bitcoin Spot ETF Sees $471.4 Million Net Inflow, Ethereum ETF $120.2 Million]On April 7, according to monitoring data from Farside, the US Bitcoin spot ETF had a net inflow of $471.4 million yesterday, while the Ethereum ETF saw a net inflow of $120.2 million.

US Bitcoin Spot ETF Sees $471.4 Million Net Inflow, Ethereum ETF $120.2 Million

On April 7, according to SoSoValue data, the US spot Bitcoin ETF had a net inflow of $471 million yesterday.

[US Spot Bitcoin ETF Sees $471 Million Net Inflow Yesterday]On April 7, according to SoSoValue data, the US spot Bitcoin ETF had a net inflow of $471 million yesterday.

US Spot Bitcoin ETF Sees $471 Million Net Inflow Yesterday

On April 7, according to SoSoValue data, the US spot Ethereum ETF experienced a net inflow of $120 million yesterday.

[US Spot Ethereum ETF Sees $120 Million Net Inflow]On April 7, according to SoSoValue data, the US spot Ethereum ETF experienced a net inflow of $120 million yesterday.

US Spot Ethereum ETF Sees $120 Million Net Inflow

According to monitoring by 1M AI News, Chris Park, head of the X developer platform, announced that the X API will continue to be free for public utility applications that provide emergency alerts and public service updates. Additionally, a new API program aimed at public utility developers will be launched, offering clear access guidelines, expanded access permissions, and more resource support. On February 6 of this year, X shifted its API pricing to a pay-per-use model, eliminating the free tier for new developers, with costs set at $0.01 per post and $0.005 per read. Previously, free access for public utility applications was granted on a case-by-case basis without a publicly available standard process. This announcement formally upgrades this exemption to a structured developer program, providing a long-awaited but clear assurance for public welfare applications that rely on X data to distribute critical information, such as emergency alerts and weather warnings.

[X Launches Public Utility Developer Program, Emergency Alert API Remains Free]According to monitoring by 1M AI News, Chris Park, head of the X developer platform, announced that the X API will continue to be free for public utility applications that provide emergency alerts and public service updates. Additionally, a new API program aimed at public utility developers will be launched, offering clear access guidelines, expanded access permissions, and more resource support. On February 6 of this year, X shifted its API pricing to a pay-per-use model, eliminating the free tier for new developers, with costs set at $0.01 per post and $0.005 per read. Previously, free access for public utility applications was granted on a case-by-case basis without a publicly available standard process. This announcement formally upgrades this exemption to a structured developer program, providing a long-awaited but clear assurance for public welfare applications that rely on X data to distribute critical information, such as emergency alerts and weather warnings.

X Launches Public Utility Developer Program, Emergency Alert API Remains Free

Bitmine disclosed an increase of 40,302 ETH in the past week, with a total holding of 4.24 million ETH, accounting for 3.52% of the circulating supply. Among them, 2.009 million ETH have been staked, and based on a 2.81% Compound Ethereum Staking Rate (CESR), the annualized yield reaches 164 million USD.

[Bitmine's Ethereum staking has reached 2 million coins, with annualized returns potentially exceeding $164 million.]Bitmine disclosed an increase of 40,302 ETH in the past week, with a total holding of 4.24 million ETH, accounting for 3.52% of the circulating supply. Among them, 2.009 million ETH have been staked, and based on a 2.81% Compound Ethereum Staking Rate (CESR), the annualized yield reaches 164 million USD.

Bitmine's Ethereum staking has reached 2 million coins, with annualized returns potentially exceeding $164 million.

Article

Santiment: Bitcoin Shows Short-Term Price Top Signal at Weekend Close

Iranian Senior Official: Iran Will Not Accept 'Diktat'

BTC broke through 70000  USDT

Trump Reiterates Threat to Attack Iranian Bridges and Power Plants

Fed's Probability of Keeping Rates Unchanged in April at 98.4%

Anthropic Expects Annual Revenue to Exceed $30 Billion by 2026

Iran Proposes Using Strait 'Toll' for Reconstruction, Sharing with Oman

Bitmine's Ethereum staking has reached 2 million coins, with annualized returns potentially exceeding $164 million.

Clause de non-responsabilité：Le contenu ci-dessus ne représente pas les positions de HTX.，HTX ne fournit pas de recommandations de trading.。

Articles connexes

Tous les commentaires0RécentPopulaire

Sujets d'actualité

Articles connexes