RT @elder_plinius: @sama

I miss @ilyasut.

literally me rn with my ability to express emotions

@NielsRogge and this is also interesting. It looks like this is the reason for the system and developer roles in the new Responses API that replaces the Completion API.

RT @pli_cachete: Getting ready

guys

Opus 4.1 failed the 3D cube test 👀

gpt-oss 20B from @OpenAI is great for tool use. this is the first open-source model I've seen that was able to complete this multi-action task. Qwen 30ba3b fails to call the first tool even in this case. However, I suspect it might be a bug related to Ollama tool use for

Alice 3.6.0 will roll out today, introducing advanced MCP tool use for OpenAI, Anthropic, Gemini, Groq, and Ollama, as well as your Custom Backend. It's a new chapter for the Alice app, which is now agentic. Stay tuned.

If you wanted to use `gpt-oss:20b` via @ollama but your initial impression wasn't great, update Ollama to the latest version and try again. Be sure to use the function-calling format. Just check out this video: .1. Fetched current crypto prices from CoinMarketCap .2. Saved the

RT @bindureddy: Opus 4.1 Barely Moves The LiveBench Needle, GPT-OSS IS MIA!. Opus 4.1 is a marginal improvement over Opus 4. Effectively, t….

RT @ivanfioravanti: So far I prefer Qwen3 over OpenAI gpt-oss 🤷🏻‍♂️, but more weeks of test drive are needed for sure.

RT @gdb: team has been working super hard, excited for tmrw!.

RT @theo: Vercel complaint of the day. It sucks that AI SDK doesn’t get first class support for OpenRouter. Seems to be due to competition….

RT @OpenAI: LIVE5TREAM THURSDAY 10AM PT.

RT @ClementDelangue: Lots of conflicting takes about gpt-oss (yay open-source in the spotlight)! . We’re powering the official @openai demo….

horizon alpha / beta is beast. It has helped me design improved agentic logic for @MCP_Community tool use, supporting parallel function calling. this video is at real-time speed. and the fact that GPT-4.1-nano handles this logic is 🔥

RT @scaling01: xAI API broke my mental. valued at $200B yet they don't even have a functional API for Grok-4. it's so fucking cringe. I'm o….

Is it just me, or is gpt-oss-120b randomly returning empty content?