GetChain News
中简 中繁 EN
GetChain News
Toggle sidebar

Online/Update

News linked to both this project and an event.

Sarvam is developing sovereign AI while also reselling DeepSeek models.

Indian AI company Sarvam has launched Token Factory, offering inference for Chinese open-source models such as DeepSeek, Kimi, GLM, and Qwen on GPUs within India. As a beneficiary of this sovereign AI initiative, the company is now caught in a debate over whether to develop its own models or act as a distributor.

B.AI Popular Model Discount Update: GLM-5.3-Flash Now Discounted to 60%

B.AI announces another upgrade to its model discount matrix. Discounts for GLM-5.3-Flash increase from 30% off to 40% off, while Hy4 is available at a limited-time 50% off, DeepSeek-V4.1-Flash at 40% off, and Qwen3.8-Flash at 20% off, covering use cases such as code development, deep reasoning, and multimodal Agents. Leveraging intelligent routing and compute aggregation capabilities, developers can seamlessly switch between major models after a single integration. Each invocation features precise routing and metering, allowing applications to be built at a more optimal cost. Visit chat.b.ai/chat to experience it.

The Ethereum project Golem raised 820,000 ETH ten years ago, and the foundation has disclosed that it has now staked 100,000 ETH

Odaily reports: Julian Zawistowski, an executive at the Ethereum project Golem Foundation, published a post reviewing the project's 10th anniversary. He said that on November 11, 2016, the Golem project launched its token sale, raising approximately 820,000 ETH in less than 30 minutes and minting 1 billion tokens, originally named GNT and now renamed GLM.Zawistowski disclosed that since 2023, the Golem Foundation has staked 100,000 ETH and distributes the staking yields to GLM holders participating through Octant, while also supporting the development of Ethereum public goods. Its long-term goal is to safeguard this reserve of funds so that GLM holders and the broader Ethereum ecosystem can benefit.

Model Service Provider Verification Service Launches, Mira Network Introduces Sterling with Public Ratings Every Two Hours

Odaily reports: On-chain analyst Ember posted on X that Mira Network has launched Sterling to verify model service providers. Sterling tests the outputs of various models and service providers every two hours and publishes ratings publicly, routing requests only to verified service providers. It currently supports Claude, GPT, and Gemini models, with DeepSeek and GLM coming soon; some models offer discounts of up to 50% below list price, and payments can be made using USDC and MIRA tokens.

Newton Foundation: Binance attacker mistakenly transferred tokens worth approximately $3.5 million into the NEWT contract, now permanently locked.

According to a post by the Newton Foundation (@NewtonFndn), the attacker involved in the recent Binance breach appears to have made an operational error, directly transferring approximately 8 million NEWT tokens, along with 17 other tokens (ANIME, ANKR, ARPA, BARD, CELR, DENT, DOLO, FTT, GLM, HIGH, HOT, KNC, LQTY, RARE, RESOLV, SIGN, WOO) to the NEWT token contract address (0xd0ec028a3d21533fdd200838f39c85b03679285d). Valued at current prices, the total amounts to approximately $3.5 million. The foundation stated that the NEWT token contract itself was not compromised. Featuring an immutable design with no owner or administrator, all management rights were relinquished on-chain in 2025, and upgrade functionality has been permanently disabled. As a result, no one, including the foundation, can retrieve these tokens.

Doubao Work launches third-party model services for enterprise customers.

Doubao Work has announced the rollout of third-party large model services for enterprise customers, initially integrating models such as Kimi K3.1 Pro, GLM-5.3, and DeepSeek V4.1 Flash. This update is available to users of the Team Subscription and Enterprise Subscription plans. The feature will be automatically enabled upon launch, requiring no separate procurement from model vendors or additional applications from enterprises. Usage will subsequently be deducted from the enterprise’s existing AI points based on actual consumption.

ZCode provides 300 million free GLM-5.3-Flash Tokens daily.

ZCode stated that the platform provides 300 million free GLM-5.3-Flash tokens to users daily. Users can claim them in-app starting at 11:00 UTC and use them from 15:00 UTC to 01:00 UTC the following day. Users must update ZCode to the latest version, switch to Global Building after claiming, and select the GLM-5.3-Flash model.

MiniMax Claims Space Bunny, M3.1 Flash Official Release Coming Soon

The official MiniMax account posted a short video captioned "Welcome back, Space Bunny." This move was viewed as an official acknowledgment of the anonymous model Space Bunny. Since its launch in late September, the model topped the daily token usage rankings on both OpenRouter and OpenCode within three days, placing among the top three alongside DeepSeek V4.1 Flash and GLM 5.3 Flash.

Security agency: Hackers used LLM combined with AI penetration tools to attack South Korean financial institutions, leaking data of approximately 68,000 people

security firm CrowdStrike released an intelligence report stating that an unidentified hacker, between late September and early October, combined large language models (LLMs) with the domestic open-source AI penetration testing tool ARTEX to launch targeted cyberattacks against multiple South Korean financial institutions and steal data.After analyzing the hacker's exposed open directory, it was found that they employed a dual-server architecture centered on a Hong Kong IP, and used models including Claude Code, GLM-5.3, Grok 4.6, and DeepSeek v4.1-flash accessed through an API proxy to orchestrate the attack process. The hacker also asked Claude how to monetize the leaked South Korean data in Telegram groups. According to estimates by South Korean media, at least seven South Korean financial institutions were attacked, including KB Kookmin Bank, Shinhan Bank, and Hana Bank, with personal data of approximately 68,000 people leaked, involving sensitive information such as annual income and loan limits.

B.AI Popular Model Perks Update: Starting September 25, MiMo-V2.6 Series Takes Over Special Offers

Starting September 25 at 15:00 (SGT), B.AI will roll out a new round of upgrades to its popular model benefits: the brand-new MiMo-V2.6-Flash takes over with 90% off pricing, while the flagship MiMo-V2.6-Pro enjoys a half-price launch special. Currently, DeepSeek-V4.1-Flash, GLM-5.3-Flash, Qwen3.8-Flash, Hy3, and MiMo-V2.5 are enjoying their final 90% discounts. Developers are urged to take advantage of this final window to top up credits and make calls in advance to ensure seamless workflow continuity. Log in to chat.b.ai/chat to experience it now.

B. AI Responses API Adds Support for GLM and Kimi Series Models

B.AI announces that its Responses API officially supports the GLM and Kimi model series. Using a B.AI API Key, developers can directly call these two model series within the Codex client. Combined with the previously supported GPT and DeepSeek series, it now covers four mainstream model lineups. After completing the streamlined configuration of the API Key and Base URL, you can seamlessly integrate the reasoning, code, and text processing capabilities of these models into your daily development workflow, flexibly meeting diverse coding and business requirements. Log in to b.ai to experience it now.

GLM-5.3-FlashX officially launches on B.AI, with simultaneous access to both the API and Web Chat.

The B.AI platform announced that Z.AI's speed-optimized native multimodal large model GLM-5.3-FlashX has officially launched, with both the API and Web Chat now open for public testing.

Zhipu officially launches GLM-5.3-FlashX

Zhipu has officially announced the launch of GLM-5.3-FlashX. The model was previously made available to global developers under the name "Ox Alpha". According to Zhipu, leveraging inference compute from 100,000 domestic chips, the company has further increased infrastructure investment and optimized inference performance, achieving speeds of up to 200 Tokens/s. The GLM-5.3-FlashX API is now live, with the model identifier set to "GLM-5.3-FlashX".

Zhipu GLM Unveils Recursive Self-Improvement Practice, GLM-5.3-Flash Released

The Zhipu GLM team has disclosed its first engineering implementation of recursive self-improvement. An Infra Agent driven by GLM-5.3 completed the design, debugging, and optimization of the GLM-5.3-Flash inference infrastructure. The practice was executed on a domestic 100,000-GPU cluster, which the team describes as the first recursive self-improvement practice for large models in China.

Just 10% of the official price: B.AI’s Top 5 Most Popular Models Now Fully Live.

B.AI officially announces the arrival of five popular models now available at just 10% of their standard price. Starting September 16 at 17:00 (SGT), Qwen3.8-Flash, Hy3, and MiMo-V2.5 officially join the 10% pricing tier, teaming up with DeepSeek-V4.1-Flash and GLM-5.3-Flash—already available at this rate—to form B.AI’s exclusive five-model discounted bundle. This means developers can seamlessly switch across diverse workloads including inference, coding, multimodal processing, Chinese-language applications, and high-frequency APIs by paying only 10% of the official standard rate, empowering them to build cost-efficient AI infrastructure. Whether for independent development or large-scale team deployments, B.AI makes top-tier compute power accessible, truly delivering a low-cost, highly flexible model calling experience. Start your trial now: https://chat.b.ai/chat

B.AI Model Benefit Update: Two Flagship Models Now Fully Live at 10% Rate

B.AI announces that a new round of benefits for its popular models is now officially in effect: exclusive 10% discounted API call privileges for high-concurrency flagship models DeepSeek-V4.1-Flash and GLM-5.3-Flash are now fully available, allowing users to invoke them at just 10% of the original price to meet high-throughput scenario demands with exceptional cost-efficiency. Simultaneously, Qwen3.8-Flash, Hy3, and MiMo-V2.5 continue to offer fully free API calls, providing developers with richer zero-cost options. This benefit adjustment delivers maximum value for high-throughput needs while leveraging a multi-model matrix to lower the exploration threshold for developers. B.AI continues to empower developers to build low-cost AI infrastructure through a diverse model lineup and inclusive pricing, eliminating prohibitive compute expenses and accelerating the rapid deployment of innovations.

GLM-5.3-Flash Official Price Adjustments Take Effect Soon; B.AI Free Tier Extended Until September 12

Zhipu's official GLM-5.3-Flash will undergo a pricing update shortly. To provide global developers with a more ample preparation window, B.AI has announced a limited-time extension of its free access—until September 12 at 09:59 (SGT), GLM-5.3-Flash will remain completely free on the B.AI platform. Leading in platform usage volume, GLM-5.3-Flash features 320B total parameters and 18B activated parameters, supports a 1M ultra-long context window, and seamlessly combines rapid response with powerful reasoning capabilities, making it a highly cost-effective choice for high-frequency coding, massive data processing, and complex Agent workflows. Additionally, other popular models such as Qwen3.8 Flash, Hy3, and MiMo V2.5 continue to be 100% free on the B.AI platform. B.AI remains committed to supporting global developers with inclusive computing resources, ensuring that frontier model capabilities are truly within everyone's reach.

OpenAI CFO: Enterprise revenue grew 32% in a single month; Luna usage surged approximately 10x after an 80% price cut

Odaily News: OpenAI CFO Sarah Friar stated that the company is accelerating the expansion of AI into specialized fields such as chip design, life sciences, and financial services, and is experimenting with pricing based on business outcomes rather than usage. OpenAI's enterprise business revenue grew 32% from June to July this year, while the company's overall annualized revenue increased approximately 20% during the same period. As of mid-year, revenue from enterprise and consumer businesses had split roughly evenly.Additionally, OpenAI recently cut the price of its low-cost Luna model by 80%, which was followed by an approximate 10-fold increase in usage. Friar noted that in cloud deployment scenarios, Luna's cost is even lower than Z.ai's GLM 5.3. OpenAI's coding tool Codex has now reached 25 million users. The company is also leveraging its own AI models to assist in developing the Jalapeno chip and completed chip design tape-out within nine months. (Reuters)

B.AI continues to provide free calls to GLM-5.3-Flash, maintaining zero-cost development.

B.AI has announced that after Zhipu's official limited-time 50% discount expires at 24:00 on September 9, GLM-5.3-Flash will continue to provide zero-cost API calls for global developers, requiring no changes to usage habits due to upstream price adjustments.

GLM-5.3-Flash Tops B.AI Model API Call Volume Rankings, Cumulative Throughput Surpasses 2.41 Trillion Tokens

The GLM-5.3-Flash model has become the most frequently invoked and popular model on the B.AI platform, with cumulative token throughput exceeding 2.41 trillion. As the first native multimodal model in the GLM-5 series, GLM-5.3-Flash features 320B total parameters and 18B active parameters. It employs a hybrid architecture combining sparse and linear attention mechanisms, supports 1M ultra-long context windows, and balances rapid response, powerful reasoning capabilities, and high cost-effectiveness. Starting today, developers can still invoke this model for free via the B.AI platform, covering diverse scenarios such as high-frequency APIs, coding, complex Agents, and ultra-long document processing. Try it now: chat.b.ai/chat