Open Weights an der Arena-Spitze. Google stockte. Anthropic gab nach.

AI Changelog · Ausgabe 17Week 29 · Jul 13–19, 2026

Moonshot AIs Kimi K3 schlug Fable 5 bei Arena Coding zu einem Drittel des Preises — das dritte chinesische Frontier-Modell in 30 Tagen — am selben Tag, an dem Google die vierte Verzoegerung von Gemini 3.5 Pro seit Mai bestaetigte. Anthropic beendete Fable 5's vierwoechige Zugangskrise mit einem dauerhaften Abo-Tier, und China lancierte ein paralleles KI-Governance-Gremium mit 29 Nationen.

266 announcements scanned · 48 mattered · 1 verdict changed

Geänderte Urteile

Pending

GA mid-late July 2026 (I/O May 2026 announcement)vorherige Wahl
Indefinite — months behind scheduleneue Wahl

Fourth postponement since May I/O announcement; Bloomberg confirms model 'months behind' on coding benchmarks with key DeepMind researchers having departed. Google preparing stopgap 'Gemini 3.6 Flash' while 3.5 Pro remains in development.

Vollständige Begründung

Vor der nächsten Ausgabe handeln

Kimi K3 Tops Arena Coding at $3/$15 — the world's largest open-weight model

Moonshot AI's 2.8T-parameter MoE scored 1679 on Arena Code WebDev, beating Fable 5 and GPT-5.6 Sol on front-end coding at one-third the price. Third Chinese frontier model in 30 days. Weights promised by Jul 27. Already paused new subscriptions due to demand.

Details & Schritte

Thinking Machines Inkling ships — 975B open-weights MoE, Apache 2.0

Mira Murati's first model: controllable thinking effort, strongest open-weight safeguards (FORTRESS 78.0%), self-fine-tuning demo in 27 minutes. Explicitly 'not the strongest overall' — positioned as a customization base, not a frontier champion.

Details & Schritte

Fable 5 gets permanent subscription access — Max/Team Premium at 50% limits

Fourth access-policy reversal in four weeks: Max ($100/mo) and Team Premium plans now include Fable 5 at 50% of limits starting Jul 20. Pro gets one-time $100 credit. Reverses the 'credit-only for everyone' plan that defined the previous three weeks.

Details & Schritte

Microsoft trains sales force to pitch against OpenAI and Anthropic

Bloomberg: internal FY27 strategy meeting coached sales teams to position MAI models as faster, cheaper, and more integrated. EVP Andreou's slide deck: Claude is 'slower and less accurate' inside Office. Third major signal of Microsoft's AI-partner decoupling in two weeks.

Details & Schritte

Fable 5 extended to July 19 — second extension in five days

Anthropic pushed the credit-cliff deadline again as GPT-5.6 Sol went GA at half Fable 5's price. The pattern of 5-7 day extensions signaled compute scarcity and competitive pressure. Superseded by the Jul 18 subscription announcement.

Details & Schritte

GPT-Red makes GPT-5.6 Sol 6x harder to jailbreak

OpenAI disclosed the first dedicated frontier-scale safety red-teaming model. GPT-Red achieved 84% attack success vs human red-teamers' 13%, and adversarially training Sol against it cut prompt injection failures 6x. Sol stays Recommended — this strengthens the existing stance.

Details & Schritte

Auf unserem Radar

  • [WATCH] Opus 5 launch window Jul 20-26 — Honeycomb EAP artifact in Cursor — Multiple independent leakers converge on this week for Opus 5 launch. Unreleased 'Claude Honeycomb EAP' briefly appeared in Cursor's model picker with 1M-token context. Fable 5 subscription resolution removes the credit-cliff-bridge narrative but timing remains aligned.

  • [WATCH] DeepSeek V4 — Jul 24 API migration deadline, $74B raise, IPO prep — Legacy endpoints retire in 4 days. V4 reportedly 'within days' of GA at Opus 4.8-level performance with $0.0028 pricing. Founder Liang Wenfeng is now the world's richest AI creator at $16.7B.

  • Gemini 3.5 Pro confirmed delayed — 'months behind' on coding — Bloomberg confirmed the fourth postponement since I/O May 2026. Internal checkpoints 'failed to match GPT-5.6.' Four senior DeepMind researchers departed. Stopgap 'Gemini 3.6 Flash' rumored. Verdict: Pending timeline hardened to 'indefinite.'

  • GPT-5.6 Sol autonomously deleted user files — OpenAI pre-disclosed risk, shipped anyway — Root cause: $HOME variable processing error expanding into recursive home-directory cleanup. At least one production DB destroyed. OpenAI's June system card warned of 'unsafe actions autonomously' — first major product-safety incident for the GPT-5.6 family.

  • Anthropic files confidential S-1 — October 2026 IPO target — Wilson Sonsini handling. $47B annualized revenue run rate. Polymarket: 78% IPO probability by Dec 31. Will be one of the largest tech offerings in history — and a watershed for AI as an asset class.

  • Nadella calls Fable 5 'editorially controlled' — Microsoft-Anthropic tension goes public — Microsoft CEO told Copilot engineers that Anthropic's refusal policy 'doesn't make sense' and described its model as 'editorially controlled.' A sitting CEO publicly attacking a partner's flagship product is rare — and comes as Microsoft's MAI strategy positions both OpenAI and Anthropic as competitors.

  • WAICO — 29 nations form Chinese-led World AI Cooperation Organization in Shanghai — Xi Jinping announced WAICO at the World AI Conference. Members include China, Russia, Pakistan, Indonesia, Venezuela, and Belarus — with UN Secretary-General Guterres attending. This is the institutional infrastructure for a parallel AI governance order outside US influence.

  • Grok Build CLI open-sourced under Apache 2.0 — new 0-day found immediately — SpaceXAI published 844K+ lines of Rust after the repo-upload scandal. SlowMist discovered a new trust-boundary bypass vulnerability one day after release. Open-sourcing doesn't resolve the trust problems when new 0-days surface immediately.

  • Claude for Teachers — free education product, Gates Foundation backed — Anthropic launched free premium access for verified K-12 educators with standards-aligned tools. Strategic land-grab in the education vertical while competitors focus on enterprise.

  • Kimi K3 pauses new subscriptions due to demand surge — splits into web/code tiers — Just 4 days after launch, demand pushed compute capacity 'close to its limits.' Echoes Fable 5's capacity struggles. Company adding capacity in batches.

  • [WATCH] Meta-Anthropic $10B compute deal in early talks — NYT exclusive: Anthropic proposed leasing Meta's AI data center compute in a potential $10B deal over two years with monthly payments and exit rights. Diversifies Anthropic's compute beyond AWS ahead of October IPO.

  • [WATCH] US considering FINRA-like independent AI model regulator — Trump administration exploring independent regulator modeled after FINRA for vetting frontier AI models. Would replace case-by-case Commerce Department review. Response to ad-hoc government interventions exposed by the Fable 5 ban cycle.

  • [WATCH] UK AISI: open-weight models now match frontier cyber from 4-7 months ago — GLM-5.2 matches Opus 4.6 on all 70 narrow cybersecurity tasks. Autonomous attacks: $1.19 on DeepSeek V4-Pro vs $85 on closed frontiers — 70x cheaper. Open-weight cyber gap narrowed from 6-10 months (2025) to 4-7 months.

  • [WATCH] xAI 2T Grok 4.6 targeting August 2026 — Musk confirmed on X: Grok 4.6, a 2-trillion-parameter successor, is in the pipeline with initial training wrapping up — just two days after Grok 4.5's public release.

  • [WATCH] Hunt.io: Claude Code + DeepSeek-v4-pro found in China-linked intrusion — First documented case of AI coding tools used as attack infrastructure in suspected state-linked espionage. Agent-governance controls could close the attack gap — AI coding agents are now confirmed as attack vectors in real intrusions.

  • Alibaba Qwen3.8 Max preview — fourth Chinese frontier model in a month — 2.4T-parameter multimodal model claiming 'second only to Fable 5.' No benchmarks or open-weight terms yet. Alibaba shares rose 5.4%. The cadence is accelerating: GLM-5.2 → Hy3 → Kimi K3 → Qwen3.8, plus Inkling from the US.

  • AI supply chain poisoning demonstrated for under $100 — Security researcher poisoned an open-weight AI model for less than $100. Supply chain vulnerability for the expanding open-weight ecosystem as Chinese and US models proliferate.

  • Grok 4.5 takes #2 on FrontierSWE leaderboard — Beats Claude Opus 4.8 and GPT-5.5 at $2/$6 pricing — competitive validation for xAI's coding claims.

Aussortiert

  • Consumer products (~40+ items): iOS 27 Siri AI public beta, Spotify AI chatbot, Roblox AI game creation, OpenAI $230 Codex keyboard, Google Vids AI avatars, Meta smart glasses, OpenAI ChatGPT speaker, Instagram AI deepfakes, consumer wearables
  • Re-reports/dedupes (~50+ items): GPT-5.6 Sol GA, Apple/OpenAI lawsuit, Fable 5 extension, Kimi K3 launch, Anthropic S-1, Nadella comments — every major story had 3-15 duplicate reports
  • Opinion/commentary/analysis (~35+ items): 'AI Isn't Human' op-eds, Nadella monopoly commentary, VC opinion pieces, AI cost-backlash essays, 'generative AI is an engineering disaster' think pieces
  • Fundraises and corporate (~25+ items): PixVerse $439M, Nous Research $1.5B, Fireworks AI $17.5B, Applied Computing $20M, Gradium $100M, Lyzr $100M agent-led raise, Vertu $6,880 AI phone
  • Show HN / personal projects (~30+ items): sqlite-utils 4.0, llm-coding-agent alpha, various vibe-coded apps, personal blog posts, developer experiments, AI Arcade
  • Personnel and governance: Johannes Heidecke departure, Ben Bernanke LTBT appointment, Microsoft AI cybersecurity layoffs, Fidji Simo OpenAI role change
  • Entertainment and niche: Lorde says AI glasses 'not sexy,' AI-generated Odyssey movie, Suno YouTube training data, Character.AI microdramas, Aries horoscope (Brave search artifact)
  • Enterprise/regional filler: Google NY educators summit, Google UK AI productivity report, Alabama/Virginia data center PR, various regional AI policy pieces
  • Financial/legal: IBM shares crash, SpaceX stock drop, OpenAI EU trademark loss, NY State data center moratorium, publisher lawsuits against Google Gemini training

Our take: Die Open-Weight-Aera ist angekommen. Wenn ein 2,8 Billionen Parameter grosses chinesisches Modell ein amerikanisches Frontier-Modell an der Coding-Spitze zu einem Drittel des Preises uebertrifft — und drei Open-Weight-Anwaerter im selben Monat Frontier-Territorium erreichten, in dem Google seine vierte Deadline verpasste — veraendert sich die Marktstruktur dauerhaft. Anthropics Fable 5-Zugangschaos war die andere Haelfte derselben Geschichte: das letzte geschlossene Frontier-Modell, das mit Compute-Knappheit kaempfte, waehrend offene Wettbewerber seinen Preisvorteil aufzehrten. Die Abo-Loesung am 18. Juli war ein wettbewerbsbedingtes Zugestaendnis im Gewand einer Produktentscheidung — Anthropic erkannte, dass $10/$50 pro Million Tokens fuer ein Modell mit schrumpfenden Faehigkeitsabstaenden die Abwanderung von Abonnenten beschleunigen wuerde. Unterdessen unterzeichneten 29 Nationen Xi Jinpings WAICO in Shanghai: die institutionelle Infrastruktur fuer eine Governance-Ordnung, die keine US-Genehmigung benoetigt. Die KI-Welt spaltete sich diese Woche in zwei — nicht durch Politik, sondern durch Produkte.

— The Neomanex Editorial Engine

Ausgabe 18 erhalten

Die Urteilsänderungen und Fristen der nächsten Woche, in deinem Postfach.

Mit dem Abonnieren stimmst du unserer Datenschutzerklärung zu. Jederzeit abbestellbar.