LIT climbs another 10% in 24 hours, lifting its market capitalization to $1.277 billion.

Beating AI News reports that Google DeepMind has launched the AlphaGenome Atlas, a searchable database precomputing all ~9 billion possible single-letter DNA variations in the human genome. The entire dataset is around 1 petabyte (PB), over 30 times larger than the AlphaFold Database. With the human genome containing roughly 3 billion DNA letters, each position can be replaced by 3 other bases, leading to the ~9 billion single-letter variations. Previously, researchers had to analyze variations individually using AlphaGenome; now they can directly retrieve results from the Atlas, avoiding repeated model runs. The Atlas also provides an AVI score for each variation, integrating predictions from AlphaGenome and AlphaMissense to help researchers prioritize variants most likely to have functional impacts. It covers both the ~2% protein-coding regions and the remaining 98% non-coding regions, which are harder to study. Multiple research teams have already used the tool to identify previously overlooked rare disease-related variations, with experiments confirming one such variant causes abnormal RNA splicing.

According to HTX market data, Bitcoin has rebounded to break through $79,000, while Ethereum has also rebounded above $2,500.

US tech journalist Mark Gurman has revealed that Apple (AAPL.O) has finalized the name of its first foldable iPhone as iPhone Duo, with a starting price of $2,000 — a figure that accounts for additional memory costs. The device will offer a maximum storage capacity of 2TB, corresponding to a price tag of around $3,000. The iPhone Duo is slated for an official launch in October, supports the Apple Pencil, and comes in two color options: dark blue and white.

Ant Bailing has officially open-sourced its first native multimodal model, Ling-3.0-flash-VL. The model was previously launched via API; this release includes BF16 and FP8 weights under the MIT license, with FP4 and INT4 versions to follow later. Ling-3.0-flash-VL can process images, videos, documents, and software interfaces, then call tools to complete operations, and further check and modify based on execution results. For example, when building a webpage, it can write code based on reference images, then review the actual page performance and adjust it on its own. The model uses a Mixture of Experts (MoE) architecture, with approximately 124 billion total parameters and around 5.5 billion activated parameters per inference run. Its context window can be extended up to 1 million tokens. Ling-3.0-flash-VL inherits the core advantage of Ling-3.0-flash as an efficient execution node in agent workflows, balancing output quality and operational efficiency in visual feedback loops, enabling it to complete full tasks at lower cost and in less time.

According to EmberCN’s monitoring, a crypto whale transferred 209.1 BTC to Binance roughly half an hour ago, worth about $16.44 million. Over the past three months, this whale has sold a total of 329.1 BTC at an average selling price of $75,392. The BTC were originally accumulated in 2022 at an average cost of $36,066, generating a total profit of approximately $12.94 million and a return of 109%.

Ant Group’s open-source announcement has officially unveiled and open-sourced Ling-3.0-flash-VL, the first native multimodal large model in its Ling series. Derived from the MoE (Mixture of Experts) architecture of Ling-3.0-flash, the model has a total parameter count of 124 billion, with 5.5 billion parameters activated per inference. It natively supports image, text, and video inputs, and features a 256K-token context window. Focused on the theme of “how to complete real-world tasks more reliably and efficiently”, Ling-3.0-flash-VL explores key directions including: Integrating visual capabilities into large models— a common concern is that this will diminish text intelligence, but training practices have yielded the opposite result: native multimodal joint training not only expands application boundaries but also enhances text intelligence. The model introduces a visual feedback mechanism: observing execution results, comparing against targets, identifying discrepancies, and iterating corrections, shifting tasks from one-off “generation” to a closed loop of “observe → act → verify → correct” to boost execution reliability. It also inherits Ling-3.0-flash’s core strength as an efficient execution node in Agent workflows, balancing output quality and operational efficiency within the visual feedback loop, enabling full task completion at lower cost and in shorter time.