Learn  /  What's new

Data platform news this week: Snowflake Expedition, Databricks metric view relationships, and Mistral Large 4 goes open

Snowflake's biggest launch since Summit, a quiet but important semantic-layer feature from Databricks, three model retirements to check, and a trillion-parameter model you can now download. The week, filtered to what changes your work.

What's new7 min readUpdated 3 November 2026
SnowflakeDatabricksDatabasesWhat's new
All guides and cheat sheets

This week belonged to Snowflake, which opened its new Expedition event on 3 November with a long list of launches. Databricks shipped smaller but useful changes to its semantic layer and MLflow, BigQuery was quiet after a busy October, and the AI model world moved again. Here is the short list from 28 October to 3 November 2026. Last week's edition covers 21 to 27 October.

25Snowflake launches and updatesat Expedition, day one
3Databricks model retirementscheck your endpoints
1Trillion parametersMistral Large 4, now open

Snowflake: Expedition 2026

Snowflake Expedition, 3 November

  1. CoCo Cloud Agents and Automations

    The coding agent runs in Snowflake and on a schedule, with restricted session scopes.

  2. Interactive Analytics and Elastic Channels

    Sub-second answers for apps and agents, and streaming that scales with volume.

  3. Snowpark Connect for Iceberg and Spark code migration

    Run Spark on Snowflake with fewer changes; CoCo rewrites Spark code.

  4. Pipeline Builder, CoCo GitHub integration, Cortex TrainingNow

    Visual pipelines, agent changes through pull requests, and post-training open models on your data.

Read the full breakdown in Snowflake Expedition 2026: every announcement that matters.

Databricks

Semantic layer

Beta
  • Relationships between metric views: declare how two metric views join, with cardinality, so AI assistants and BI tools stop guessing from column names
  • Declared with SQL, for example ALTER VIEW ... ADD CONSTRAINT ... MANY TO ONE RELATIONSHIP TO ...
  • Not enforced and not used for query optimisation; metadata only

A small feature with a big effect on AI answer quality.

MLflow and governance

Beta and GA
  • Managed warehouse for MLflow traces (beta): view traces in the MLflow UI without choosing a SQL warehouse
  • Higher limits and BROWSE permission for Unity Catalog secrets
  • Track Unity Gateway skill usage in audit logs and the usage dashboard

Easier monitoring of AI apps and agents.

Models and connectors

Plan ahead
  • Microsoft Graph Logs connector in beta for Lakeflow Connect
  • Retired: Moonshot AI Kimi K2.7, Thinking Machines Inkling and DeepSeek V4 Pro (0813)

Update any jobs or dashboards that call the retired endpoints.

SQL
-- Tell AI assistants and BI tools how orders relate to customers
ALTER VIEW main.sales.orders_mv ADD CONSTRAINT orders_customer
  MANY TO ONE RELATIONSHIP TO main.sales.customers_mv
  USING (customer_id);

BigQuery

No new BigQuery release notes appeared between 28 October and 3 November, after almost 30 in October. If you missed them, our BigQuery October 2026 update covers reusable column aliases, AI.AGG and TimesFM 3.0, and our new BigQuery vs Snowflake comparison puts them in context.

AI models that matter for data teams

  • Mistral Large 4 open weights (29 October): about 1 trillion parameters, available on Hugging Face in FP8, NVFP4 and GGUF versions. The licence allows commercial use except selling it as a hosted model service. API price: $0.70 input and $2.10 output per million tokens. See Mistral Large 4, explained.
  • Gemini 4 Carbon (28 October): first on the Artificial Analysis Intelligence Index v5 with 59.4. See Gemini 4 Carbon, explained.
  • A tougher leaderboard: Artificial Analysis launched version 5 of its index with mostly private tests. See the best AI models in November 2026.
  • GitHub Octoverse 2026: agents now start over 40% of pull requests. See what it means for data teams.

Check your understanding

1/3What do Databricks relationships between metric views do?

2/3Which Snowflake feature for real-time answers is generally available?

3/3What should you do about the Databricks model retirements?

More platform news: last week's data platform news, Snowflake vs Databricks for AI agents and Microsoft Fabric vs Databricks.

Sources: Snowflake's Expedition 2026 press release and blog (3 November 2026); Databricks platform release notes (late October and November 2026) and documentation on metric view relationships and the MLflow managed warehouse; BigQuery release notes; Hugging Face and Mistral AI on Mistral Large 4 (29 October 2026). Features roll out by region and edition.

Alessandro Ecclesie AgazziWritten by Alessandro Ecclesie Agazzi, analytics engineer in London. Updated 3 November 2026. About me

Found this useful?

New guides and "what's new" explainers on data and AI land here regularly. Follow along on LinkedIn, or share this one with a colleague.