OpenAI
GPT-5.4
- OSWorld-Verified (OpenAI)
- 75 %
OpenAI releases GPT-5.4 (새 탭에서 열림)
The mini and nano tiers land at roughly four times the GPT-5 prices, so re-check cost models before switching defaults.
GPT-5.4 with computer use, Claude drives the desktop, auto mode instead of permission prompts, and Mistral Small 4 under Apache 2.0.
OpenAI
GPT-5.4
The mini and nano tiers land at roughly four times the GPT-5 prices, so re-check cost models before switching defaults.
Apache 2.0 on vLLM, llama.cpp, and SGLang, so one model replaces three in a self-hosted stack.
gemini-3-pro-preview was shut down on March 9 and now aliases to 3.1 Pro, which can silently change pinned pipelines.
Simplifies RAG stacks that currently glue separate text and image embedders together.
A voice-first agent no longer needs a speech-to-text and text-to-speech sandwich.
Where a connector exists it uses that instead of the screen, so the fragile path is the fallback.
Directly relevant if you run with permissions skipped today or maintain allowlists by hand.
Proposals along these four lanes move fastest, so aim contributions at them.
Admins get private plugin repos with governance, which is what makes editor plugins adoptable at work.
A concrete template for delegating experiment loops to agents.
Useful for anyone building their own agent loops instead of consuming finished ones.
The open-weights camp is moving up the stack, from models to platforms for building them.