Skip to content

Su

Business coverage of technology, AI and cryptocurrency

Last update

Search results for «LLM Optimization»

9 posts

21 October 2

  1. DeepSeek DeepInfer Open Source Release Cuts LLM Inference Costs

    DeepSeek released its DeepInfer inference framework under Apache 2.0 on October 20, 2026, targeting lower costs for large-scale LLM serving.

    2 min

  2. MiniMax M3.1 Flash Launches with 70% Lower Agentic AI Costs

    MiniMax released its M3.1 Flash model on October 19, 2026, claiming major gains in speed and efficiency for agentic AI applications at flat API rates of $0.15 input and $0.60 output per million tokens.

    3 min

6 November 1

  1. Together AI Compiler Achieves Up to 2.6x Inference Throughput on Frontier Models

    Together AI introduced its agent-driven compiler on November 4, 2026, delivering measurable gains in inference speed across multiple hardware platforms and frontier open models.

    3 min

5 November 1

  1. NVIDIA AI Agents Context Window Study Finds 16K Tokens Maintain Accuracy

    A new NVIDIA preprint demonstrates that frontier models retain performance on agent tasks with frequent compaction to 16K-token windows instead of full long contexts.

    3 min

4 November 1

  1. Mistral Large 4 Open Weights Released for 1-Trillion-Parameter Le Chonk Model

    Mistral AI released open weights for its 1-trillion-parameter Mistral Large 4 model on October 29, 2026, enabling self-hosted coding agents and multimodal applications.

    3 min

29 October 1

  1. Google Cloud AI platform Unveils Ironwood TPU for Enterprise Inference

    Google Cloud launched Ironwood, its seventh-generation TPU, delivering substantial inference cost reductions and performance gains for enterprise AI deployments.

    3 min

19 October 1

  1. Meta Muse Spark Open 816B Model Released for Coding and Agentic Tasks

    Meta released Muse Spark Open, an 816-billion-parameter open-weight multimodal model designed for complex coding and agentic tasks, on October 15, 2026.

    3 min

17 October 2

  1. Ant Group Ling Model Open Weights Enable 560B MoE Deployment on Four GPUs

    Ant Group’s inclusionAI published open weights for Ling-3.1-Flash, a 560 billion parameter MoE model under MIT license, on Hugging Face in mid-October 2026.

    3 min

  2. Vitalik Buterin AI privacy stack adds Tor, zkAPI and local prompt rewriting

    Vitalik Buterin released an open-source privacy wrapper for his personal AI agent on October 16, 2026. The package routes traffic through Tor, uses zero-knowledge payments and rewrites prompts locally before external model calls.

    3 min