Baseten

using GLM-5.3 to summarize a webinar about itself?? this week, we hosted a webinar that well over 1k people were interested in, including many engineers and leaders at F500 cos. It was co-led by Philip, author of Inference Engineering, and Declan, MOTS at Artificial Intelligence. They talked all things GLM-5.3 from benchmark results to what it takes to running it in production. I decided to use GLM-5.3 on Baseten MAPI to summarize the transcript from the recording and Inkling to "watch" the YouTube upload and find some pull quotes. all for $0.25.

We’re thrilled to welcome Paul Sinai and the Blaxel team to Baseten. As agents become central to how software gets built and used, the infrastructure stack has to expand beyond inference. Blaxel brings that execution layer to the platform teams already trust for training and inference. Owned intelligence is going to need a new infrastructure stack, we’re building one together. https://lnkd.in/gyrsVhdW

We are looking for (another) amazing executive recruiter to help us fill the growing number of leadership roles across Baseten. Two profiles that could be particularly strong: a talent partner at a VC firm, or an executive recruiter from a top agency. The common thread is someone who has spent years building and nurturing relationships with A+ talent and understands that the best leadership hires are not transactional. If you’re interested in the role, or know someone who should be, please email Lou Zumpano at lou.zumpano@baseten.co.

in 2 weeks i'll be in portugal for a buddy's wedding. inspired, i one-shotted a website for a 3-day festival in lisbon. model: GLM-5.3 provider: baseten tokens: 3734 cost: $0.017 temperature: the sun

Baseten was built on the belief that the health of the open-source AI ecosystem matters. Base Labs is our commitment to that. Our engineers are doing frontier research on continual learning, RL, and model performance, and sharing every experiment and recipe openly along the way.

The sales leadership team at Baseten is hardcore, smart, and obsessive. The group also happens to be very fun and, apparently, very tall (esp in my presence). Join us: https://lnkd.in/gADdEhfJ

What better way to end the week than with a shoutout( https://lnkd.in/gi6DnCdX ) from the vLLM team for our work optimizing Qwen3-TTS on top of vLLM-Omni. Kudos to Ian Carrasco for driving this work. At Baseten , we’re big believers in the open-source community. We’re committed to helping customers run the best open-source models in production with the highest performance, best reliability, and lowest cost. In our latest technical blog post, we share how we optimized Qwen3-TTS for high-fidelity voice agents and voice cloning, bringing serving costs down to around $3 per 1M characters — about 90% lower than comparable closed-source alternatives — while keeping performance faster than real time. Blog post: https://lnkd.in/g_pvGFhZ We are #hiring ! If you love building AI products, enjoy hard technical challenges, believe in the future of inference, and are excited about voice technology, reach out! Job post: https://lnkd.in/eygwTTZ2 #AI #Inference #OpenSourceAI #VoiceAI #TextToSpeech #TTS #vLLM #Qwen #LLMServing #AIInfrastructure #Baseten #Hiring

It's been awesome to see the impact our customer's products are having across so many different industries. More models = more options for AI teams to build with and I'm excited to see what hard problems Baseten customers will conquer next.

In the words of Baseten CEO, Tuhin Srivastava, "Intelligence should be defined by the teams closest to the work. Intelligence should be owned by all of us." Shoutout to Parallel Web Systems , OpenEvidence , Braintrust and many others who are owning their intelligence. We're proud to power the inference layer for the teams that are shaping the future 💚