When trying to understand GPT-OSS, you must consider price:performance.

Lots of evals in the GPT-5 release! We'll independently verify as many as possible and add them to OpenBench 🫡

PSA: We just shipped a fix to improve tool calling in multi-turn conversations for the GPT-OSS models via Groq API. 🛠️ You can now pass `reasoning` back to the chat completions endpoint in the messages array to maintain context and properly pass tool calls back to the models.

👀

If you used the GPT-OSS models in the last couple days for tool calling on Groq Inc, give them another shot!

Now it's even MORE stacked with one more capability: Structured Outputs

👑 The Browsers OSS King has been dethroned 👑 Anchor Browser has integrated with Groq to provide the state-of-the-art Browser Agents API, based on the fastest GPT-OSS API. A huge step up in: 🚀 High agentic accuracy 🚀 Blazing fast agentic loop

We just hit 2M devs. Intern Dylan Mitic can stay.

Groq just hit 2M devs. Threw a party…

And we’re just getting started! Join us!

this is my official petition to replace our intern @dylanmitic with one of our 2M developers that actually possess PHD level intelligence to avoid further chart crimes

Reminder that Groq API supports structured outputs for GPT-OSS, Kimi K2, and Llama 4 models to guarantee responses strictly conform to your JSON schema. We've noticed a lot of you using `json_object` and highly recommend using `json_schema` instead. 🫡

Because Groq Inc is so fast, generating dynamic webpages on the fly with GPT-OSS feels like an index.html

Quick tip: You can configure OpenAI's Codex to use Groq Inc as a model provider for GPT-OSS, Kimi K2, and any other models you like coding with via Groq API. ⚡️ Same tool, faster inference, and much cheaper. Here's how to set it up:

17 new benchmarks in OpenBench! Easy to setup and get reproducible benchmarks across any provider and model.

SemiAnalysis It’s unfortunate that we can’t trust model makers’ eval results. This is an issue we’re trying to solve with OpenBench - an easy standard way to run evals easily transparently. We just released v0.2.0 and we’ll be adding SWE-Bench-Verified shortly 🫡 github.com/groq/openbench

⚡ Generate lightning-fast subtitles that transcribe + translate videos in seconds w/ Groq Inc + FFmpeg. Upload a video, pick the origin and translation language (50+ supported), and get burned-in subtitles to make your content accessible worldwide. 🧵

Announcing Open Lovable 🔥 We've built an open-source AI web app builder that can transform any website URL into a working, editable clone, giving you a foundation to build on instantly. All powered by Groq Inc, E2B, and Firecrawl.

Congrats to Recall for hitting #1 on Product Hunt this week powered in part by Groq Inc and Compound Beta