AI Newswire
Wire bulletin — Thu, 01 Oct 2026 00:18:48 UTC
All times UTC

MODELS DESK

Frontier-model launches, benchmark results and technical disclosures.

23 filings on the wire.

  1. [MODELS] Filed Sep 10, 2026 UTC

    SAN FRANCISCO, September 3 — OpenAI begins rolling out GPT-6 Astra computer-use agent to paid ChatGPT tiers

    By Mateo Quintana, Wire correspondent

  2. [MODELS] Filed Sep 04, 2026 UTC

    MENLO PARK, September 4 — Google's AI Max auto-migration of small-business Search campaigns begins Sept. 1

    By Mateo Quintana, Wire correspondent

  3. [MODELS] Filed Aug 19, 2026 UTC

    SAN FRANCISCO, August 18 — OpenAI launches age-gated ChatGPT for Teens with content blocks and parental controls

    By Mateo Quintana, Wire correspondent

  4. [MODELS] Filed Aug 19, 2026 UTC

    SAN FRANCISCO, August 18 — OpenAI halts largest frontier training run after Astra nears critical cyber tier

    By Mateo Quintana, Wire correspondent

  5. [MODELS] Filed Aug 18, 2026 UTC

    SAN FRANCISCO, August 18 — OpenAI launches ChatGPT for Teens with study mode, parental controls amid youth-safety lawsuits

    By Mateo Quintana, Wire correspondent

  6. [MODELS] Filed Aug 16, 2026 UTC

    HANGZHOU, August 15 — Alibaba open-sources Qwen3.8-27B under Apache 2.0 as Qwen tops 3 billion downloads

    By Mateo Quintana, Wire correspondent

  7. [MODELS] Filed Aug 15, 2026 UTC

    HANGZHOU, August 15 — Alibaba releases Qwen3.8-27B open weights under Apache 2.0

    By Mateo Quintana, Wire correspondent

  8. [MODELS] Filed Aug 15, 2026 UTC

    MENLO PARK, August 15 — Alibaba publishes Qwen3.8-27B open weights, model tops Hacker News

    By Mateo Quintana, Wire correspondent

  9. [MODELS] Filed Aug 14, 2026 UTC

    MENLO PARK, August 13 — Google ships Gemini 3.7 Flash three weeks after 3.6, halves token pricing, still no date for 3.5 Pro

    By Mateo Quintana, Wire correspondent

  10. [MODELS] Filed Aug 13, 2026 UTC

    MENLO PARK, August 12 — SpaceXAI ships Grok 4.6, matching GPT-5.6 Sol at roughly 60% lower cost

    By Mateo Quintana, Wire correspondent

  11. [MODELS] Filed Aug 10, 2026 UTC

    SAN FRANCISCO, August 10 — OpenAI files GPT-5.6-Cyber, splits Daybreak into two tiers as it pauses more powerful Astra

    By Mateo Quintana, Wire correspondent

  12. [MODELS] Filed Aug 09, 2026 UTC

    SAN FRANCISCO, August 7 — OpenAI flags Astra as first model that may reach 'Critical' cyber threshold, pauses parts of development

    By Mateo Quintana, Wire correspondent

  13. [MODELS] Filed Aug 08, 2026 UTC

    SAN FRANCISCO, August 7 — OpenAI cannot rule out 'critical' cyber capability in unreleased Astra model, pauses some internal work

    By Mateo Quintana, Wire correspondent

  14. [MODELS] Filed Aug 04, 2026 UTC

    SHANGHAI, August 4 — MiniMax releases H3 video model weights but bars US, EU, UK and South Korea from local deployment

    By Mateo Quintana, Wire correspondent

  15. [MODELS] Filed Jul 27, 2026 UTC LEAD STORY

    BEIJING, July 27 — Moonshot AI releases full weights of 2.8-trillion-parameter Kimi K3 model

    By Mateo Quintana, Wire correspondent

  16. [MODELS] Filed Jul 26, 2026 UTC

    SAN FRANCISCO, July 24 — Anthropic launches Claude Opus 5, pitching frontier-adjacent performance at half of Fable 5's price

    By Mateo Quintana, Wire correspondent

  17. [MODELS] Filed Jul 20, 2026 UTC

    SHANGHAI, July 20 — Moonshot's Kimi K3 triggers second 'DeepSeek moment' as chip stocks tumble

    By Mateo Quintana, Wire correspondent

  18. [MODELS] Filed Jul 10, 2026 UTC

    MENLO PARK, July 8 — SpaceXAI launches Grok 4.5, first model since IPO and Cursor deal

    By Mateo Quintana, Wire correspondent

  19. [MODELS] Filed Jul 01, 2026 UTC

    SAN FRANCISCO, June 30 — Anthropic launches Claude Sonnet 5 as default model, undercutting flagship pricing ahead of IPO

    By Mateo Quintana, Wire correspondent

  20. [MODELS] Filed Jun 29, 2026 UTC

    SAN FRANCISCO, June 26 — OpenAI launches GPT-5.6 under U.S. government-managed access list

    By Mateo Quintana, Wire correspondent

  21. [MODELS] Filed Jun 25, 2026 UTC

    SAN FRANCISCO, June 24 — OpenAI, Broadcom unveil Jalapeño, first custom inference chip aimed at cutting AI compute costs

    By Mateo Quintana, Wire correspondent

  22. [MODELS] Filed Jun 10, 2026 UTC

    SAN FRANCISCO, June 9 — Anthropic releases Claude Fable 5, first general-availability Mythos-class model, with mandatory retention and routing safeguards

    By Mateo Quintana, Wire correspondent

  23. [MODELS] Filed May 05, 2026 UTC

    SAN FRANCISCO, May 5 — OpenAI ships GPT-5.5 Instant as ChatGPT default; claims 52.5% fewer hallucinations

    By Mateo Quintana, Wire correspondent