The day in AI

Wednesday 30 September 2026

←→

Google released Gemini 4 Argon to cyber defenders first; OpenAI disrupted a distillation campaign

The day in brief

Google released Gemini 4 Argon first to vetted cyber defenders, the third lab to stage a top model's release around cyber risk, and OpenAI said it disrupted a campaign to copy its models' reasoning.

Gemini 4 Argon has a one-million-token output limit and a reported 77.9% on DeepSWE v1.1. Its first users are vetted cyber defenders, under the US voluntary pre-release access process. Developers, enterprises and paying consumers come later. Anthropic staged Mythos this way in April, and OpenAI did the same with GPT-6 Astra in September.

OpenAI said it disrupted a campaign to extract protected reasoning from its models. The campaign involved about 16,000 requests from more than 4,000 users in a two-day spike in July, and OpenAI attributed a core cluster to people associated with Moonshot AI. Read with yesterday's GLM-5.3 study, the day shows labs gating their strongest models while others close the gap fast.

Also in the news

  • SynthID Bio from Google DeepMind is a proof of concept for watermarking AI-designed proteins in the sequence itself.
  • Cohere Embed 5 (Pro and Fast) produces multimodal embeddings with a 128K context.

Everything from this day

Launches and products

30 Sep 2026Google DeepMindGemini 4 Argon Gemini 4 Argon, a new frontier model with a 1M-token output limit, scores 77.9% on DeepSWE v1.1; released first to vetted cyber defenders via Fairwind.30 Sep 2026Synopsys / OpenAIGPT-Synopsys Synopsys and OpenAI announced a multi-year partnership to build GPT-Synopsys, a specialized model that operates Synopsys EDA tools to run chip design workflows, with shared revenue.30 Sep 2026CohereEmbed 5 (Pro and Fast) Embed 5 is Cohere's newest embedding family, in Pro and Fast tiers, with multimodal inputs and 128K context.

Research

30 Sep 2026OpenAIDisrupting a coordinated model-distillation campaign OpenAI says it disrupted a campaign extracting protected reasoning from its models and attributes a core cluster to people associated with Moonshot AI.30 Sep 2026Google DeepMindSynthID Bio SynthID Bio is a proof of concept for watermarking AI-generated proteins in the sequence itself while preserving function in lab tests.