What was announced at the OpenAI DevDay 2023 keynote?

GPT-4 Turbo (Preview), Price cut for GPT-4 Turbo (GA), Copyright Shield (GA), Whisper v3 (GA), Custom Models program (Announced), GPTs (GA), and 1 more.

What were the highlights of the OpenAI DevDay 2023 keynote?

OpenAI's first developer conference launched GPT-4 Turbo with a 128K context and a much lower price, new modalities in the API, Copyright Shield, custom GPTs that anyone can build without code, and the Assistants API for building agent-like experiences. Microsoft CEO Satya Nadella joined Altman on stage. GPT-4 Turbo supports up to 128,000 tokens of context, 16 times GPT-4's 8K, with knowledge updated to April 2023. DALL-E 3, GPT-4 Turbo with vision and a new text-to-speech model all came to the API. Copyright Shield: OpenAI will defend customers and pay the costs of copyright infringement claims.

Who gave the OpenAI DevDay 2023 keynote?

Sam Altman, on 2023-11-06 in San Francisco, California.

OpenAI DevDay 2023: Opening keynote

Sam Altman · Nov 6, 2023 · 46m · San Francisco, California

Watch on YouTube

OpenAI's first developer conference launched GPT-4 Turbo with a 128K context and a much lower price, new modalities in the API, Copyright Shield, custom GPTs that anyone can build without code, and the Assistants API for building agent-like experiences. Microsoft CEO Satya Nadella joined Altman on stage.

Key themes: Models, Agents, Pricing · AI-written summary from the published transcript

Where each highlight is said45:36

1 2 3 4 5 6

6:05
GPT-4 Turbo supports up to 128,000 tokens of context, 16 times GPT-4's 8K, with knowledge updated to April 2023
8:05
DALL-E 3, GPT-4 Turbo with vision and a new text-to-speech model all came to the API
11:44
Copyright Shield: OpenAI will defend customers and pay the costs of copyright infringement claims
12:35
GPT-4 Turbo is 3x cheaper than GPT-4 for prompt tokens
31:03
GPTs, custom versions of ChatGPT, can be built without code; a GPT Store was promised for later that month
43:35
The Assistants API makes it easier to build assistant experiences with threads, retrieval, Code Interpreter and function calling