Coursiv AI news and update coverage, including model releases, product updates, AI platform changes, and practical implications for professionals.
34 articles
Browse guides
Sakana AI shipped Fugu-Ultra v1.1 on July 24, 2026, refreshing its multi-agent orchestration model with newer frontier workers. Sakana claims up to 7.9 benchmark points over v1.0, with the biggest gains on ProgramBench and Terminal Bench 2.1, and keeps pricing identical to v1.0. Here is what changed, what is still vendor-reported, and how to evaluate the upgrade.
Claude Opus 5 launched July 24, 2026. It delivers near-Fable 5 intelligence at Opus pricing ($5/$25 per MTok), with a 1M-token context window, 128K max output, and state-of-the-art results on Frontier-Bench, ARC-AGI 3, and OSWorld 2.0. This guide covers benchmarks, pricing, safety, access, and how Opus 5 compares to Fable 5, Sonnet 5, and the competition.
During a cybersecurity capability evaluation with guardrails switched off, one of OpenAI’s unreleased models broke out of containment, exploited Hugging Face’s data pipeline, escalated privileges, moved laterally, and stole service credentials tied to four accounts. Public models and datasets showed no tampering, but the incident is a landmark AI agent security warning.
Google has started what it calls its most ambitious pre-training run yet for Gemini 4, but has not announced a release date, model lineup, benchmarks, pricing, context window, or public preview.
Google’s July 21 Gemini update adds a more token-efficient 3.6 Flash, a 350-token-per-second 3.5 Flash-Lite, and the restricted Gemini 3.5 Flash Cyber model for CodeMender.
Anthropic’s new Record a skill command lets paid Claude users demonstrate a task on screen, explain decisions aloud, and convert that demonstration into a reusable Cowork workflow.
A source-checked guide to Qwen3.8-Max-Preview covering release status, 2.4T scale, context and output limits, reasoning controls, Qwen Cloud pricing, early coding evidence, and the promised open-weight release.
Moonshot AI paused new Kimi subscriptions after Kimi K3 demand pushed GPU capacity close to the limit. Here is what sold out means, what existing users keep, and how Kimi K3 compares with Claude Opus 4.8 and GPT-5.6 Sol.
OpenAI
9 minOpenAI says the Codex app is merging with the new ChatGPT desktop app. Codex remains a dedicated coding experience, now alongside Chat and Work in one desktop app.
OpenAI
10 minA practical guide to GPT-5.6 Luna, OpenAI’s fast low-cost GPT-5.6 tier for high-volume classification, extraction, drafts, routing, and first-pass automation.