Z.ai

Flexible pay-as-you-go pricing for Z.ai models and services.

Dario Amodei argues that frontier AI labs should deliberately slow the pace of model capability improvements. The article frames this proposal against Beijing’s speed-first AI safety regime and describes AI safety as an urgent debate.

Oct 2026 · SemiAnalysisNews article

Anthropic launches Claude Haiku 5.5, its first new version of its smallest model in nearly a year, with input and output prices for requests under 100,000 tokens reduced to $0.10 and $0.50 per million tokens. The company estimates average savings of about 75%; Haiku 5.5 also adds effort controls and improves on Haiku 4.5 across reported benchmarks, including computer use and agentic coding. The model is available through the Claude Platform, AWS, Google Cloud, and Microsoft Azure.

Oct 2026 · The New StackNews article

Mistral unveils Mistral Large 4, a one-trillion-parameter model it says ranks among the world’s strongest open models, with strengths in cyber, coding, manufacturing, finance, and multimodal tasks. Trained for two months on 4,000 Nvidia Grace Blackwell GPUs in Mistral’s European data centers, it launches in preview to developers, cybersecurity leaders, and state authorities before a wider release later in October. Mistral says the model will rank among the top open-weight systems on aggregate benchmarks, though it still trails frontier models in areas such as coding.

Oct 2026 · CNBC TechNews article

SemiAnalysis measures usage limits across AI subscription plans and introduces a dashboard tracking plans from Anthropic, OpenAI, Meta, SpaceXAI, Cursor, Cognition, Z.ai, MiniMax, and Moonshot. Its rough estimate says Anthropic subscriptions generate 10% of revenue while consuming more than 40% of inference compute, lowering blended revenue per megawatt by about $36 million. Experiments isolate input, cache-write, cache-read, and output tokens to estimate how plans use limits across models.

Oct 2026 · SemiAnalysisNews article