Baseten

Try DeepSeek V4 Flash on our Model APIs today. - 80–98% cheaper than other frontier models - Comparable intelligence - 1M token context window Use it here: https://lnkd.in/dDKZnhAD

Open-source TTS models like Qwen3-TTS are very useful out of the box supporting zero-shot voice cloning. However, the ability to fine-tune them on your use-case specific voices makes them significantly more expressive and life-like. We published a modified Qwen3-TTS recipe in our ml-cookbook to help users perform voice cloning via supervised fine-tuning. With this recipe, we see 16% faster time to first audio (TTFA) against other cloning techniques while unlocking a path to further control emotion and prosody. See how it's done below: https://lnkd.in/eqXYjy69

We're excited to welcome Mani Parkhe to our engineering team! Mani joins us with 25+ years of experience at companies like Databricks, Uber, LinkedIn, Mentor Graphics, and Cadence Design Systems. Beyond technical depth, Mani embodies the positive, human culture of any company he joins. We are fully aligned that company culture matters as much as technical acumen, and we couldn't be happier to have him on board. We're growing fast. If you want to join Mani in working on problems that touch some of the most ambitious companies in the world, we're hiring!

Thinking Machines Lab’s new Inkling-Small is now available on Baseten! At 276B parameters (12B active), Inkling-Small retains the powerful multimodal capabilities and 1M-token context window of the original Inkling (975B, 41B active), but with a significantly smaller footprint. Try it now on our Model APIs: https://lnkd.in/dFEgmwmY

We believe in a future where open-weight models coexist with dozens of specialized closed-weight models. Today, we're launching Baseten for Model Labs to continue to serve that future. Baseten for Model Labs provides the infrastructure, distribution, and go-to-market support so labs can continue to focus on building great models.

Over the past year, we've all been amazed by what new models let us do. Not just the largest models built for large-scale reasoning and coding, but models from smaller, specialized labs solving specific problems for businesses. We believe this diversity of labs and models is vital to a healthy, growing ecosystem — one where companies building with AI can draw on a wide variety of models to build the best possible products and experiences. The future of AI is not a handful of general-purpose frontier models; it's a diverse ecosystem of open and closed models, each the best in the world at its domain. That future needs infrastructure to distribute, meter, and monetize all of them in one place. We've seen this firsthand from our own customers, who ask for more models from smaller labs in their product workflows on Baseten. So… Today we're launching Baseten for Model Labs 🎉 Baseten for Model Labs is a distribution platform for closed-weight and restricted-license models: a lab's model is listed in the Baseten Model Library, and developers use it under their existing Baseten contract. No new signup, no new sub-processor, and your weights stay fully protected. We're launching with 15 incredible partners: Bria AI, Canopy Labs, Cartesia, Cosine, Gradium, Inception, krea.ai, MongoDB, Musubi, NVIDIA, pyannoteAI, Scaled Cognition, SID.ai, Subconscious, and Synthefy If you're a lab with a closed model and you want distribution without building the infrastructure: let's talk. Read the full announcement: https://lnkd.in/eXUUmQht Checkout the models here: https://lnkd.in/e924YdJG

Today, we're announcing Baseten for Model Labs. We believe the AI landscape will be made up of a diverse ecosystem of specialized, closed-weight models running alongside open-weight ones. Baseten for Model Labs is the platform built to enable labs to monetize, distribute, and scale their closed models on Baseten infrastructure. Huge shoutout to our incredible launch partners: Bria AI, Cartesia, Canopy Labs, Cosine, Gradium, Inception, krea.ai, MongoDB, Musubi, NVIDIA, Poolside, pyannoteAI, Scaled Cognition, SID.ai, Subconscious, Synthefy, and Trajectory. Learn more in our launch blog: https://lnkd.in/eXJxxVjk

Jeanne DeWitt Grosser - the COO of Vercel and former CBO of Stripe - is one of the best executives I know. She is smart, strategic, operational, and builds from first principles. This Thursday, I'll sit down with her to talk about her approach to scaling revenue, from sales segmentation to the role of AI in her organization, to unique viewpoints on hiring. Join us for the live conversation at Baseten HQ.

Happening today! Link to register in thread.