Morning Brief: October 20, 2026 — 5 AI developments worth knowing

MiniMax and DeepSeek update their model offerings, Gemini adds BigQuery analysis and classroom insights, and GitHub opens an API for Copilot model policies.

View sources
In this article
  1. 1. MiniMax launches M3.1 Flash with lower token prices
  2. 2. DeepSeek updates Flash and opens web search APIs
  3. 3. Gemini queries BigQuery data from Google Sheets
  4. 4. Classroom adds summaries of assigned Notebook conversations
  5. 5. GitHub opens an API for Copilot model access policies
  6. Sources & method

The morning view

Hosted model access, connected data permissions and administrative controls determine which of these changes readers can use now.

1. MiniMax launches M3.1 Flash with lower token prices

Development:

MiniMax rolled out M3.1 Flash on October 19 through its API, M Plan subscription and MiniMax Code. The model targets coding and workflows that use multiple tools. Published API rates are $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens.

For requests containing up to 512,000 input tokens, M3.1 Flash’s input and output rates are half M3’s standard rates. MiniMax’s claims of faster responses and stronger task performance remain vendor results. Downloadable model weights remain a future release.

Our read: Has teeth. Published per-token rates give developers a concrete cost comparison for agent workflows that make repeated model calls.

Source 1Source 2

2. DeepSeek updates Flash and opens web search APIs

Development:

DeepSeek released V4.1 Flash 1017 on October 17 across its website, app and API, alongside public Search and Fetch interfaces. Developers can use their existing DeepSeek API key to find web pages and retrieve page text for an agent. The company reports improved project-building performance for the updated model.

Search and Fetch each cost $0.75 per 1,000 successful requests. Their server-side tool calls through DeepSeek Harness, Claude Code and Codex are temporarily free for DeepSeek API users; billing begins October 31 at 16:00 UTC. Fetch remains subject to websites’ availability and access policies.

Our read: Has teeth. Documented search and page-retrieval endpoints let developers add live web evidence to agent workflows.

Source 3Source 4

3. Gemini queries BigQuery data from Google Sheets

Development:

Google began rolling out Gemini analysis of BigQuery, its cloud data warehouse, in Connected Sheets on October 19. Gemini queries the full connected dataset, beyond the cells visible in a spreadsheet. Users can review a proposed plan before Gemini builds a tab containing refreshable formulas, pivot tables and charts.

Users need permission to query the connected data. Eligible subscriptions include Business and Enterprise Standard or Plus, Google AI Pro or Ultra, Google AI Pro for Education, and AI Expanded Access. The rollout may take up to 15 days, so eligible accounts may not see the feature immediately.

Our read: Has teeth. Connected, refreshable outputs give analysts a report they can revisit as source data changes.

Source 5

4. Classroom adds summaries of assigned Notebook conversations

Development:

Google started rolling out Gemini Notebook insight summaries in Classroom on October 19. Teachers can see engagement summaries, recurring questions and areas where students may need support. Individual chat transcripts are available so teachers can check the AI-generated interpretation against the conversations.

The feature covers notebooks assigned through Classroom, with a notice telling students their teacher can see those chats. It requires administrator enablement and student activity in an assigned notebook. Education Fundamentals, Standard and Plus receive the feature over a one-to-three-day rollout; support for Gemini’s separate LTI learning-platform integration is promised in the coming weeks.

Our read: Has teeth. Recurring questions can help teachers decide what to revisit in class, while transcripts let them check the AI’s interpretation.

Source 6

5. GitHub opens an API for Copilot model access policies

Development:

GitHub made its Copilot model-policy API generally available on October 19. Administrators can retrieve and change policies for enterprises, enterprise teams, agent apps and organizations. This provides a programmable way to manage which models people can use across a company.

GitHub’s enterprise guidance covers Copilot Business and Enterprise plans. Although the API is generally available, granular access through enterprise teams remains an opt-in preview. In that mode, organization-level model settings are deactivated, and a team cannot enable a model the enterprise has disabled.

Our read: Has teeth. Central administrators can automate model access changes while retaining enterprise-wide restrictions.

Source 7Source 8

Sources & method

AI-assisted morning roundup based on primary announcements and documentation opened for this edition. Assessments are editorial judgments about documented capabilities and access, not endorsements or hands-on test results. Vendor performance claims have not been independently tested.

  1. MiniMax: M3.1 Flash release announcementPublished October 19, 2026
  2. MiniMax API Docs: Pricing OverviewAccessed October 20, 2026
  3. DeepSeek: V4.1 Flash 1017 release announcementPublished October 17, 2026
  4. DeepSeek API Docs: Search & FetchAccessed October 20, 2026
  5. Google Workspace: Analyze BigQuery data using natural language with Gemini in Google SheetsPublished October 19, 2026
  6. Google Workspace: Classroom Notebook insight summaries and sharing controlsPublished October 19, 2026
  7. GitHub: Copilot model policy API for enterprise, orgs, and teamsPublished October 19, 2026
  8. GitHub Docs: Managing availability of models in your enterpriseAccessed October 20, 2026

Source pages reviewed October 20, 2026. Analysis and suggested evaluation methods are AI Toolcast’s own.

This AI-assisted, source-reported roundup contains editorial assessments, no sponsored placement and no affiliate links. Products were not tested for this edition.

Found a possible error? How to report a correction.