Elena Marsh

Covers AI model releases, AI lab news, benchmark results and the status of AI services. Works from the labs' own announcements, model cards, system cards, technical reports and documentation, and from benchmark maintainers' published results. A pen name of The Model Press research desk.

Latest

AI benchmark news ·

Claude Opus 5.5 tops the Intelligence Index on September 22, 2026

On September 22, 2026, Artificial Analysis scored Claude Opus 5.5 at 58 on its Intelligence Index at max effort.

Typographic cover reading "OCR 4.0 and Leanstral 1.5" with concentric rings on a dark brown background
AI model news ·

Mistral OCR 4.0 retired with Leanstral 1.5 on September 30, 2026

Mistral AI deprecated OCR 4.0 and the experimental Leanstral 1.5 on September 29, 2026 and retired both the next day.

Typographic cover reading "Transcribe 1.0 retired" with concentric rings on a dark violet background
AI model news ·

Grok Voice Transcribe 1.0 retired on October 2, 2026: calls move to 2.0

xAI's speech-to-text model grok-voice-transcribe-1.0 reached end of life on October 2, 2026.

AI benchmark news ·

ARC Prize publishes GPT-6 Astra ARC-AGI-3 results on September 3, 2026

On September 3, 2026, ARC Prize reported GPT-6 Astra at 62.7% on ARC-AGI-3 Semi-Private with its Standard harness.

AI model news ·

Claude Fable 5.1 and Mythos 5.1 release on September 1, 2026

Claude Fable 5.1 release: Anthropic shipped Fable 5.1 and Mythos 5.1, one model with two safeguard levels, on Sep 1.

Typographic cover reading "GPT-Synopsys, a model for chip design" on a dark amber background with abstract circuit-style lines
AI lab news ·

OpenAI and Synopsys announce GPT-Synopsys for chip design, Sep 30, 2026

OpenAI and Synopsys agreed on September 30, 2026 to build GPT-Synopsys, a model that runs Synopsys chip design tools.

AI model news ·

OpenAI model deprecations, October 1, 2026: GPT-5.1, Codex and TTS

OpenAI model deprecations on October 1, 2026 cover GPT-5.3-Codex, GPT-5.1, GPT-5.4-Nano and three TTS models.

AI model news ·

OpenAI GPT-6 Astra release on September 3, 2026: what's new

GPT-6 Astra release: OpenAI shipped its most capable model on September 3, 2026, with a Critical cybersecurity rating.

AI model news ·

AI news September 2026: models, chips and data centers

AI news for September 2026: GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon, plus chip roadmaps and power deals.

AI model news ·

AI news August 2026: models, chips and data centers

AI news for August 2026: a Google DeepMind leadership change, OpenAI's 8 GW PORTS-Pike deal and Nvidia's Q2 results.

AI model news ·

AI news July 2026: models, chips and data centers

AI news for July 2026: GPT-5.6, AMD's Instinct MI400 Series, Claude Opus 5 and new gigawatt-scale campuses.

AI conferences and events ·

OpenAI DevDay 2026, September 29: GPT-6.1 Sol, Dots and Agents API

OpenAI DevDay 2026 on September 29 brought GPT-6.1 Sol, always-on Dots agents and an Agents API with Computer Use.

AI model news ·

OpenAI releases GPT-6 Sol and GPT-6 Luna on September 22, 2026

OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, 2026 in ChatGPT Work, Codex and the API.

AI lab news ·

OpenAI disrupts Moonshot AI-linked distillation campaign, Sep 30, 2026

On September 30, 2026, OpenAI said it disrupted a distillation campaign and tied a core cluster to Moonshot AI.

AI lab news ·

OpenAI outlines safety cases for frontier AI training on Sept 28, 2026

On September 28, 2026, OpenAI said frontier AI training runs should need a written safety case before they continue.

AI model news ·

Anthropic Claude Opus 5.5 release on September 22, 2026: what's new

Claude Opus 5.5 release: Anthropic shipped the model on September 22, 2026, at Fable 5.1 level on most work.

AI model news ·

Anthropic Claude Sonnet 5.5 release on September 28, 2026: what's new

Claude Sonnet 5.5 release: Anthropic shipped the first Sonnet with cyber safeguards on September 28, 2026.

AI model news ·

Google Gemini 4 Argon release on September 30, 2026: what's new

Gemini 4 Argon release: Google shipped it on September 30, 2026 to cyber defenders first, with a 1M-token output limit.

The Model Press

What are you looking for?

Search by headline, topic or keyword.