The AI Catchup - October 13, 2026

Google released Gemini 4 Argon this week. The items that matter:

• Gemini 4 Argon is out: Google's new frontier model, first on the Vals Index at 68.9%, with up to 1M tokens of output
• ChatGPT Intelligent UI: GPT-6 reaches the Chat tab, from Free to Enterprise, with answers that can include charts, forms, and buttons
• Claude Dashboards and Motion: live data dashboards and MP4-ready animations as Claude artifacts; Docs, Slides, and Design leave beta
• OpenAI Decisions API: typed yes/no, choice, and score answers about 10x faster than the Responses API, at $0.10 per million input tokens
• The catch: Argon's API price doubles to $4 and $20 on December 1, and it is not on the free API tier or the Google AI Pro plan

Every claim below is fact-checked in the linked write-ups.

Gemini 4 Argon Is Out: Google's New Frontier Model

Before: Argon went only to trusted cyber defenders after its September 30 announcement. Now: paid API developers, Ultra subscribers in Antigravity, and Pro and Ultra subscribers in the Gemini app can use it.

• Model ID: gemini-4-argon, with a 1,048,576-token window shared by input and output
• Long decode: up to 1M output tokens through the Interactions API in background mode; 65,536 by default
• Cost: $2 input and $10 output per million tokens through November 30; $4 and $20 from December 1
• Who gets it: paid Gemini API in AI Studio and Android Studio; Antigravity desktop and CLI on Google AI Ultra; Gemini app and AI Mode on Pro and Ultra
• Coding: 77.9% on DeepSWE v1.1, a long-horizon software engineering benchmark, by Google's count
• Watch out: Claude Opus 5.5 still leads Terminal-Bench 4.0 in Vals AI's runs, and the model card warns of occasional slowness or timeouts

Access by plan, the price table, long decode setup, and when to pick it over Claude.

ChatGPT Intelligent UI: Answers You Can Click

Before: GPT-6 was not in ChatGPT's Chat tab. Now: it powers Chat on every tier, and its answers can mix in charts, forms, tappable buttons, and interactive tools when the question calls for it.

• Who gets it: Plus, Pro, Business, and Enterprise first; Free and Go from October 8
• Models: GPT-6 Sol on paid tiers, GPT-6 Luna on Free and Go
• Watch out: Work and Codex models do not change, and the Pro reasoning option does not support it

What Intelligent UI renders and where it stops.

Claude Dashboards and Motion: Two New Artifact Types

Before: Claude's artifact templates were Docs, Slides, and Design, all in beta. Now: Dashboards and Motion join them, and Docs, Slides, and Design are out of beta on every plan, including Free.

• Dashboards: beta on paid plans; ask questions of Snowflake, BigQuery, Databricks, and other sources, and each chart shows its query
• Motion: beta on Team and Enterprise; editable code-based animations that export as MP4
• Watch out: Motion is not a video generation model and makes no generated footage

Data sources, export, and what to try first.

OpenAI Decisions API: Fast, Typed Answers

Before: a yes/no or scoring check meant a full Responses API call. Now: POST /v1/decisions returns a probability, a choice, or a rubric score, about 10x faster by OpenAI's measure.

• Model: gpt-6-luna, the only one available today
• Cost: $0.10 per million input tokens; no output, cache-read, or cache-write charges
• Watch out: public beta; send dependent questions as separate requests

Request shape, question types, and pricing details.

Quick Hits

• Codex composer predictions: the Codex desktop app suggests your next message on personal Pro plans; Tab accepts it for editing, and it does not count toward usage limits during beta
• OpenAI Agents API: pre-warm a hosted sandbox with packages and files before the session starts; free until attached, expires after five minutes unattached

Ship It This Week

1. Run your biggest refactor or long-horizon coding task on gemini-4-argon before December 1, while it costs half.
2. Set background: true and a max_output_tokens above 65,536 for any Argon job that needs long output.
3. Move yes/no and rubric checks from the Responses API to /v1/decisions and compare latency.
4. Point a Claude Dashboard at one warehouse table you report on weekly.

Account and Billing Notes

• Cursor's X Subscription: Cursor now offers plans from $8 to $200 a month that pool Cursor and Grok usage weekly; existing subscribers keep their plan unless they switch, and unifying accounts is permanent. Details.
• Claude artifact promotion: discounted five-hour usage for artifacts ends October 15. Details.

---

Gemini 4 Argon is the release to test this week: a new frontier model with a 1M-token output budget, at half price until December. If a teammate runs every long coding job on one model by habit, forward this.

Until next week, stay caught up.