Articles written by Diederik Knol
Fastino releases GLiNER2.5-Decide
Fastino released GLiNER2.5-Decide on 24 September 2026, a 340M-parameter open-weight model for classifying and routing text.
Agent compliance drift: What it is and how to control
An AI agent approved at go-live slowly stops matching its own documentation. The model, tools, permissions, memory and human oversight all move. This guide explains what causes agent compliance drift,

NVIDIA releases Nemotron 3 Diarization and NV-Reason-CT
NVIDIA published Nemotron 3 Diarization on 23 September 2026, a 100 million parameter open model that identifies who spoke when. NVIDIA also released NV-Reason-CT for 3D CT scans the same
Ant Group releases Ming-Image-0.1-Design
Ant Group released Ming-Image-0.1-Design and Ming-Image-0.1-Design-Layer on 22 September 2026, two 6 billion parameter image models under the MIT licence.
Xiaomi releases MiMo-V2.6-Pro and Flash
On 21 September 2026 Xiaomi released MiMo-V2.6-Pro and MiMo-V2.6-Flash under an MIT licence, with weights on Hugging Face. Pro holds 1.02 trillion parameters, 42 billion active, a 1 million token
Qwen releases Qwen-Image-2.1
On 20 September 2026 Qwen released Qwen-Image-2.1, an image generation and editing model with 7 billion parameters in its generator. It scores 60.28 on Qwen's own ranking, ahead of every
OpenAI announces Astra for Law
OpenAI unveiled Astra for Law on 17 September 2026: GPT-6 Astra combined with a US legal search index of more than 230 million URLs and instructions for legal analysis. It
Google launches Gemini 3.8 Live and Extended Thinking
Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on 15 September 2026. Extended Thinking tops the Artificial Analysis Speech to Speech Index at 82.6, just ahead of
Ant Group expands SingProbe guardrail to 29 open models
Ant Group published 27 new SingProbe checkpoints on 14 September 2026, extending its token-level safety guardrail to 29 open-weight models including DeepSeek, Qwen, GLM, Llama and gpt-oss. Each probe is
OpenAI launches GPT-Live-1 in the API
OpenAI released GPT-Live-1 in its API on 10 September 2026 at $0.05 per minute. The voice model listens and speaks at the same time, so callers can interrupt without breaking
Suno launches v6, v6-wild and v6-mini
Suno launched v6 on 9 September 2026, three music models built with Warner Music Group, BMG and Believe. v6-mini is free for all users, while v6 and v6-wild need a
OpenAI launches ChatGPT Images 2.5
OpenAI released ChatGPT Images 2.5 on 8 September 2026, an image generation and editing model it says cuts the wait for an image by up to 50% against Images 2.0.
