PulseAugur / Brief
EN
LIVE 08:23:19

Brief

last 24h
[2/2] 221 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. TOOL · Mastodon — fosstodon.org Italiano(IT) ·

    🧠 Test on an agentic task: # Qwen 3.7 Max beats # GPT 5.5 and # Claude Opus 4.7. ‼️ NO, it's not a game of tetris between the models.. 👉 Details: https://www.lin

    Qwen 3.5 Max has reportedly outperformed GPT-4.5 and Claude Opus 4.7 on an agentic task. This evaluation suggests Qwen's capabilities in complex reasoning and task execution are advancing rapidly. The specific details of the agentic task and the evaluation methodology are not fully disclosed in the provided information. AI

    IMPACT This benchmark suggests Qwen's growing competitiveness against leading models, potentially influencing future model development and adoption.

  2. RESEARCH · Mastodon — sigmoid.social English(EN) ·

    DeepSeek made its V4 Pro discount permanent Saturday: $0.87/M output tokens, 34.5x cheaper than GPT-4.5. The math holds because they run on domestic Huawei chip

    DeepSeek has permanently reduced the price of its V4 Pro model, making it significantly cheaper than competitors like GPT-4.5. This pricing strategy is enabled by their use of domestic Huawei chips for operation. The new pricing makes V4 Pro approximately 34.5 times more cost-effective for output tokens. AI

    IMPACT This permanent price cut by DeepSeek, enabled by domestic chip usage, intensifies competition and could drive down costs for AI model deployment.