Skip to the page
Tecyfy

12 October 2026 to 18 October 2026

Claude Haiku 5.5 arrives and Gemini 4 reaches the app

ReleaseDeepSeek

DeepSeek updates V4.1 Flash and opens a search API

On 17 October DeepSeek moved deepseek-flash to a new build, DeepSeek-V4.1-Flash-1017, in the API, website and app. Prices and the way you call it stay the same, but answers may change, so rerun your checks. DeepSeek also opened a public Search API on its own search engine.

Where it came from

FeatureOpenAI

GPT-Live voices can be made from a written description

Since 16 October you can describe a voice in text and OpenAI builds it, through the API or the voice Playground, then reuse it in GPT-Live sessions. Making the voice is free; speaking with it costs the normal session price. OpenAI also added 9 voices in German, French, Hindi, Japanese and Korean.

Where it came from

FeatureAnthropic

Fast mode for Claude Opus 5.5 no longer needs a request

On 15 October fast mode for Claude Opus 5.5 left research preview. You no longer ask for access or send the fast-mode-2026-02-01 beta header, and the Start, Build and Scale tiers get a default fast mode rate limit. Fast mode on Opus 5 and Opus 4.8 is still a preview, so keep the header there.

The idea behind it

FeatureDatabricks

Databricks converts T-SQL and Snowflake SQL with an agent

The agentic code converter became generally available for SQL on 15 October. Built on Genie Code, it rewrites T-SQL, Snowflake and other dialects into Databricks SQL, checks the result and fixes its own errors before saving it to a workspace folder. A warehouse migration can start from the converter's output rather than a hand rewrite, but read what it changed before you run it on real data.

FeatureOpenAI

OpenAI's GPT-Live can act while the user is still talking

From 15 October GPT-Live sessions can call Decisions directly, so an app can take an action mid-sentence or react to what the voice says. The trigger can be native delegation or the transcript as it streams in. A voice assistant no longer has to wait for the user to finish before it looks something up.

Where it came from

FeatureAnthropic

Claude API requests can now be traced with OpenTelemetry

From 14 October Anthropic exports traces of Messages API requests and Claude Managed AgentsAgentA model that uses tools in a loop, deciding each next step from what the last one returned, until the task is done. sessions to any OpenTelemetry backend, in beta. The Claude SDKs emit the spans in 7 languages, and spans from the API itself need a destination set in the Claude Console. If you already watch your services in Grafana, Datadog or Honeycomb, your agent runs can sit beside them instead of in a separate log.

The idea behind it

ReleaseGoogle

Google releases Gemini 4 Argon

gemini-4-argon became generally available on 13 October, the first Gemini 4 model in Google's API notes. The same day it reached the Gemini app for Google AI Pro and Ultra subscribers. Free accounts are not mentioned.

ReleaseDatabricks

Databricks marks Supervisor Agent as legacy

In October Databricks marked Knowledge Assistant and Supervisor Agent as legacy, 8 months after Supervisor Agent became generally available. It also published a guide to building agentsAgentA model that uses tools in a loop, deciding each next step from what the last one returned, until the task is done. in code with Agent Bricks.

The idea behind it

FeatureHugging Face

Hugging Face tokens can create repos without wider access

From 12 October a fine-grained Hugging Face token can be given a contributor permission. It can create repositories and write to the ones it made, without write access to the other repositories around them. A script that uploads models no longer needs a token that can change everything.