Together AI

Victor M

@victormustar

Open source MUST win 🔥 GLM-5.2 is free when used with Hugging Face Inference Providers and for every available provider for the next 6 hours (Zai, Together AI, Novita, Fireworks, DeepInfra) the cost is on us. Set it up with Pi, opencode, Codex, Claude Code or any coding agent to understand why people are saying open source has caught up 🔥

merve

@mervenoyann

Kimi K3 in @huggingface Inference Providers is live via @togethercompute $3/M input tokens, $15/M output tokens with 54 TPS chef's kiss https://t.co/mqNQQjiBuW

Luke Sophinos

@lukesophinos

We had Jake Saper of Emergence Capital on the podcast last week. He's backed some of the most successful vertical software companies of the last decade: Zoom, Veeva, Ironclad, and Together AI. He makes a sharp argument: AI native services businesses might have better data moats than agentic software companies selling into the same verticals. The reason is the eval loop. When you own the full outcome, you capture feedback at every layer, internal QA, customer validation, audit defense. That continuous eval data tightens the AI in ways a software tool never can, because the software vendor's view ends at the workflow. @jakesaper @picnoulos

Jordan Windmueller

@jhwindmueller

Working with Joon, Percy, Michael, Lainie, + the Simile team is the privilege of a career. Today they announce their Series B, led by Greenoaks, at a $2B valuation. We wanted to share more about why we’re such deep believers in them + what they're building https://t.co/AXb5zxeK8i

Milk Road AI

@MilkRoadAI

Chamath Palihapitiya is warning that restricting open source AI would create a cost gap so large it could ripple through the entire stock market, not just AI companies (Save this). @chamath actual argument centers on real pricing data, since he's pointed out that closing off open source access could force American companies to pay somewhere between $26 and $56 per million tokens for the same intelligence that competitors abroad can access for $0.50 to $1 per million tokens through open alternatives. That's a gap of roughly 30 to over 100 times, and his point is that this kind of cost disadvantage becomes unsustainable the moment AI turns into a core input for everyday corporate operations rather than a niche R&D expense. This debate is playing out in real time inside the industry, since real inference pricing data shows just how cheap open source models already are to run. Providers like DeepInfra are pricing certain open models as low as $0.039 per million input tokens, while Groq, Together AI and Fireworks cluster around $0.15 input and $0.60 output on the same models. However, if open-source AI does keep surviving and scaling the way it has, here is are how you can benefit from this. DeepInfra, Novita, and SiliconFlow currently lead on raw price efficiency for open source inference, and while these are largely private or foreign based, their pricing pressure benefits any public company that routes workloads through the cheapest available open model infrastructure. Nebius benefits structurally too, since its cloud infrastructure is explicitly built around running open source model catalogs at scale with cost predictability, positioning it to capture demand from companies trying to avoid expensive closed-model contracts. Meta remains a pivotal name to watch, since its Llama ecosystem is still the largest open weight foundation in the industry even as the company hedges with closed models. Nvidia also benefits regardless of which side wins, since it's been expanding its own open source push with a reported $26 billion investment and its Nemotron model line. Milk Road Pro is tracking all the biggest beneficiaries of open source AI, if you want access to all our AI trades around this trend, you can come join us for just $1 using the link below!

Pukerainbow 🤮🌈

@pukerrainbrow

Fireworks AI just raised $1.5B at a $17.5B valuation. Over $1B in revenue already. 40 trillion tokens a day. 95% of that traffic is models fine-tuned on a customer's own data. Not frontier API calls off the shelf. Together AI raised $800M days earlier. Bookings past $1.15B. Open-weight usage tripled. Yea so it turns out, nobody wants to rent intelligence anymore. They want to own it, and pay someone else to run it.

Tri Dao

Co-founder & Chief Scientist

Princeton professor and creator of FlashAttention and the Mamba architecture; serves as Together AI's Chief Scientist.

Percy Liang

Co-founder

Stanford associate professor and director of the Center for Research on Foundation Models (CRFM).