Skip to content
DigIntoAI/Model intelligence
Model intelligence feed

Model intelligence.

Follow the changes. Understand what comes next.

Editor’s pickOpenAI
OpenAI/GPT-6.2 Astra / Sol

OpenAI Launches GPT-6.2 Astra and GPT-6.2 Sol Across ChatGPT Work, Codex, and the API

At 01:45:00 Asia/Shanghai on November 18, 2026 (OpenAI News RSS; confirmed on @OpenAI at 01:54:03 Asia/Shanghai), OpenAI launched flagship GPT-6.2 Astra (gpt-6.2-astra) and everyday GPT-6.2 Sol (gpt-6.2-sol) on ChatGPT Work, Codex, and the API, bringing Ultrafast to Pro 200 starting with Sol, plus the gpt-6.2-astra-law configuration.

Latest stories485

Newest published
385

Google Open-Sources TorchTPU So PyTorch Runs Natively on TPUs

At PyTorch Conference 2026, Google open-sourced TorchTPU with Meta, a native PyTorch stack for Google TPUs. Teams can point models at a tpu device and use familiar tools such as vLLM, SGLang, and Hugging Face Transformers.

Google / TorchTPU / PyTorch / TPU / Meta / open source / vLLM / SGLang / Hugging FaceRead story
386

ChatGPT Adds Locked Chats Behind a PIN or Biometrics

ChatGPT Locked Chats move selected conversations out of regular history and search. Unlock with a PIN on web and desktop, or a PIN, Face ID, Touch ID or device unlock on supported mobile devices. Resetting the PIN permanently deletes locked chats; changing the PIN does not.

ChatGPT / Locked Chats / PIN / biometrics / privacy / OpenAIRead story
387

OpenAI Is Buying Orchestra Bio to Fold It Into Rosalind

OpenAI is acquiring Orchestra Bio, a drug-development coordination platform, and bringing the team and technology into Rosalind alongside GPT-Rosalind and Rosalind Workbench. The deal amount was not disclosed.

OpenAI / Orchestra Bio / Rosalind / GPT-Rosalind / life sciences / biopharma / acquisitionRead story
388

OpenAI Adds Optional 12-Hour Prompt Caching on the Responses API

OpenAI's Responses API can now keep a prompt cache for 12 hours. Cache writes cost 2.5× input for that lifetime, versus 1.25× for the default 30 minutes, and cache hits still use the lower cached-input rate.

OpenAI / Prompt Caching / Responses API / API pricing / GPT-5.6 / cache TTLRead story
389

Antigravity Will Remove Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b on November 2

Google's Antigravity Models page marks Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b as removed on November 2, 2026. Gemini 3.1 Pro, 3.7 Flash and 3.6 Flash are listed for Enterprise only, while Gemini 3.8 Flash is on every plan.

Google AntigravityRead story
390

Perplexity Computer Can Now Run Fully Locally on a Mac

Perplexity Computer now runs both its agent and its model on a Mac. Local tasks use PPLX Pro and cost no Computer credits, and Hybrid splits complex work between cloud and local models.

Perplexity ComputerRead story
391

Mistral Launches Voxtral 3 Speech Models With Transcription From $0.0035 a Minute

Voxtral 3 adds three transcription models, a text-to-speech model and a LiveKit-based voice agent SDK. Prices run from $0.0035 a minute, and the announcement does not mention open weights.

Voxtral 3Read story
392

Ornith Releases Ornith-1.5-122B With Open Weights Under the MIT License

Ornith-1.5-122B is a 122B MoE reasoning model with MIT-licensed open weights, a 262,144-token context window and an Ollama build. Its benchmark scores are Ornith's own.

Ornith-1.5-122BRead story
393

Here's How Gemini 4 Argon Usage Is Counted in the Gemini App, Antigravity and the API in October 2026

Google doesn't publish a set number of Gemini 4 Argon prompts for any plan. In the Gemini App, usage is compute-based and refreshes every five hours until a weekly limit. In Antigravity, Google AI Pro gets a separate weekly-only Argon quota and Ultra gets the highest limits, with no published figures. On the Gemini API, Argon is paid-only, and limits depend on the project's usage tier. This guide, based on Google pages as of 00:40 Beijing time on October 21, explains each one and what to do when you reach a limit.

Gemini 4 Argon, Gemini App, Google Antigravity, Gemini API, Google AI Studio, Google AI plansRead story
394

OpenAI Adds Team Tasks to ChatGPT So Teams Can Automate Recurring Work

OpenAI has added Team Tasks to ChatGPT, letting teams run recurring shared work on a schedule or when an event occurs, such as a new Slack message. Tasks run in the cloud with the team's service account, use workspace credits, and are controlled by enterprise admins through permissions, according to OpenAI's documentation.

ChatGPT Team TasksRead story
395

Gemini 4 Argon Costs Half as Much as Claude Opus 5.5 Until November 30, and Then the Prices Match

Google's Gemini 4 Argon and Anthropic's Claude Opus 5.5 are each company's recommended model for hard coding and knowledge work, and both have a context window of about one million tokens. Until November 30, 2026, Argon's introductory API price of $2 per million input tokens and $10 per million output tokens is half of Opus 5.5's $4 and $20. From December 1, Google's pricing page lists Argon at $4 and $20, the same as Opus 5.5. Compares price, context, access, and published benchmarks. Figures as of 22:52 Beijing time on October 20, 2026, per official pricing pages and docs.

Gemini 4 Argon, Claude Opus 5.5Read story
396

Gemini CLI Counts Requests per Day, and Here's How Many Each Sign-In Method Gets

Gemini CLI limits how many model requests you can make each day, and the number depends on how you sign in. Google's Gemini CLI quotas page as of 22:42 Beijing time on October 20, lists 1,000 requests a day for a free Google account, 250 for a free Gemini API key, 1,500 for Google AI Pro and Gemini Code Assist Standard, and 2,000 for Google AI Ultra, Gemini Code Assist Enterprise and Workspace AI Ultra. This guide covers how those limits are counted, what Gemini CLI does when you reach one, and the ways Google documents to get more.

Gemini CLIRead story

From the Webmaster

Personal experiences, thoughts and notes.

View all notes