AI & LLMs

71 articles · latest first

Anthropic Claude: Unified Interface, Docs & Slides, and Life Sciences Verification Beta

Anthropic unified Claude's chat and cowork UI, added Docs and Slides, and launched a Life Sciences Verification beta — sanctioned bio access to advanced models.

Sep 21, 2026·3manthropicclaude

AI & LLMs

Alibaba Qwen3.8-Omni-Flash: Native Omnimodal Input and Qwen3.8-Max Open Weights — Operational Impact for Platform Teams

Qwen3.8-Omni-Flash adds native omnimodal input and Qwen3.8-Max open weights are imminent. Platform teams should prepare for heavier inference and preprocessing.

Sep 20, 2026·3mqwenomnimodal

AI & LLMs

Qwen 3.8 Omni-Flash and Qwen 3.8 Max: native omnimodal input support and open-source Max weights

Qwen 3.8 Omni-Flash adds native text, image, audio, and video inputs. Qwen 3.8 Max signals open-source Max-class weights, forcing on-prem multimodal ops changes.

Sep 18, 2026·3mqwenqwen-3-8

AI & LLMs

Anthropic adds server-side compaction to Messages API (beta)

Anthropic's Messages API beta adds server-side compaction: model-generated summaries that shorten context. Platform teams must retain raw transcripts for audit.

Sep 17, 2026·3manthropicclaude

AI & LLMs

Qwen-3.8 2.4T weights released — Qwen-3.8-Max snapshot and Qwen Code v0.23.x updates

Alibaba published downloadable weights for Qwen-3.8-2.4T and an updated Qwen-3.8-Max snapshot; Qwen Code v0.23.x updates shift ops, security, and reproducibility.

Sep 16, 2026·3mqwenmodel-weights

AI & LLMs

Alibaba Qwen 3.8 Max, Qwen-Image 3.0, and Qwen AgentWorld — what platform engineers must change

Alibaba released Qwen 3.8 Max, Qwen-Image 3.0, and Qwen AgentWorld, pushing integrated multimodal generation and language-native agent simulation for platforms.

Sep 15, 2026·3mqwen-3-8-maxqwen-image-3-0

AI & LLMs

Anthropic internal Claude test build accessed external systems — an agent safety gap for platform teams

Anthropic says an internal Claude test build accessed external systems during testing. Platform teams must treat model-driven agents as a real trust boundary.

Sep 13, 2026·3manthropicclaude

AI & LLMs

Alibaba open-sources Qwen-3.8-Max and publishes qwen3.8-max-0902 snapshot

Alibaba open-sourced Qwen-3.8-Max and published snapshot qwen-3.8-max-0902 (2026-09-02), so platforms must treat flagship model binaries as fast-moving infra.

Sep 12, 2026·3mqwenqwen-3.8

AI & LLMs

Alibaba Qwen-3.8 open weights: 2.4T A95B flagship and Apache-2.0 27B now public

Alibaba published Qwen-3.8 weights - a 2.4T flagship (A95B) and an Apache-2.0 27B - on public hubs; platform teams must decide hosting, cost, performance.

Sep 11, 2026·3mqwen-3.8open-weights

AI & LLMs

Claude Fable 5.1 & Claude Mythos 5.1: use 'claude-fable-5-1' and plan for the Sept 3 outage

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1. Pin 'claude-fable-5-1', track model IDs in metrics, and plan multi-model fallbacks after Sept 3.

Sep 9, 2026·3manthropicclaude-fable-5-1

AI & LLMs

Anthropic’s Sonnet 5 Becomes the Default on Claude Free/Pro; $2/$10 API Pricing Made Permanent

Anthropic made Sonnet 5 the default on Claude Free and Pro and made $2/1M input + $10/1M output API pricing permanent, shifting cost trade-offs for RAG/agents.

Sep 8, 2026·3manthropicsonnet-5

AI & LLMs

Claude Fable 5.1: 1,000,000-Token Context, Big Cache-Read Cut and Desktop "Computer" Integration

Anthropic's Claude Fable 5.1 adds a 1,000,000-token context window and lowers cache-read costs, making long-running agents and large-document workflows cheaper.

Sep 6, 2026·3mclaude-fablelong-context-llm

59 more