Large Language Model News

Researchers from an undisclosed institution introduced AlignAtt4LLM, the first application of the AlignAtt policy to decoder-only large language models for simultaneous speech translation at IWSLT 2026. The system, a syn…

Researchers evaluated four state-of-the-art zero-shot cross-lingual voice cloning systems across six languages, finding that Arabic remains particularly challenging. The study, presented at IWSLT 2026, analyzed tradeoffs…

Researchers presented AURA-ST, a three-stage modular pipeline for low-resource speech-to-text translation at IWSLT 2026. The system bypasses traditional cross-attention by treating projected acoustic representations as t…

Barcelona Supercomputing Center submitted an end-to-end system combining a speech encoder with a translation-oriented large language model to the instruction following track of IWSLT 2026. The system uses a chain-of-thou…

Researchers from the Karlsruhe Institute of Technology developed Diet-KIT, a post-training quantization system that compresses the Qwen2-Audio-7B speech LLM from 16 GB to 3.98 GB while maintaining translation quality. Th…

HW-TSC submitted a cascaded system to the IWSLT 2026 Subtitling track, using a large-model-based streaming speech recognition framework with VAD, sliding-window context caching, long audio chunking, and Qwen3 forced alig…

A developer built a Claude Code plugin called codebase-mapper that automatically generates and maintains Markdown documentation of a project's codebase, injecting a compact index into every prompt to prevent Claude from …

Anthropic released Claude Sonnet 5 via API on June 30, 2026, as confirmed by the official inference endpoint. No benchmark scores or community sentiment have been published yet.

Anthropic released Claude Sonnet 5, its newest AI model, now available in Conductor. The model offers performance close to Opus 4.8 at a significantly lower cost.

Anthropic, AWS Bedrock, and OpenRouter all offer the same Claude models at identical per-token prices, but differ in billing, freshness, and operational features. AWS Bedrock is best for committed AWS spend or credits, A…

Andrej Karpathy's AI coding rules, The Batch's 'loop engineering' issue, and Claude Code lead Boris Cherny's flat org structure signal that the unit of AI work has shifted from prompts to self-correcting loops. This shif…

An individual has publicly declared a personal policy against using large language models or generative AI for prose, code, or any other tasks, citing a preference for human skills and interaction. The statement outlines…

The CUHKSZ team submitted a simultaneous speech translation system to IWSLT 2026, built on Qwen3-Omni-30B-A3B with LoRA adaptation, achieving 40.5 BLEU for English→Chinese and 27.7 BLEU for English→German in the 0–2 s la…

Researchers from IIIT-BGP presented low-resource Bhojpuri-Hindi speech translation systems at IWSLT 2026, including an end-to-end model combining Wav2Vec2 and NLLB-200 with a lightweight adapter, and a cascaded system us…

Anthropic released Claude Sonnet 5 on 30 June 2026, claiming it nearly matches Opus 4.8 in agentic performance at a lower price. The model is the default on Free and Pro plans, with introductory pricing of $2 per million…

Google DeepMind released the Nano Banana 2 Lite model via API on June 29, 2026, with Gemini Omni Flash integration. The model is now available for developers to build with, though no benchmark scores or community ratings…

Google DeepMind released Gemini Omni Flash on May 19, 2026, via API. The model is available but has no published benchmarks or community ratings yet.

Anthropic released Claude Sonnet 5, a mid-tier AI model that excels at writing code and handling complex tasks through deeper reasoning and self-improvement, but catches fewer bugs than earlier models and costs slightly …

Amazon released its Nova family of models on December 3, 2024, with capabilities for fine-tuning on tasks like email data extraction. The models have benchmark scores on Humanity's Last Exam and LM Arena, and the platfor…

Anthropic's Claude models became generally available on Microsoft Foundry, running on NVIDIA GB300 Blackwell Ultra GPUs, as part of a strategic partnership involving a $30 billion Azure compute commitment. The launch mak…