Together, which aims to build open-source LLMs and was part of RedPajama, a project where six companies replicated Meta's LLaMA dataset, raised a $20M seed
Victor Dey / VentureBeat :
Context & Ripple Effects
Two months after Meta stood by its release strategy even as LLaMA was leaked as a torrent, Together is capitalizing on the appetite that leak exposed: it just closed a $20M seed to build open-source LLMs, having already helped replicate Meta's training data in RedPajama, where six companies rebuilt the dataset from public sources.
The timing matters because Together isn't alone at the gate — sources report OpenAI is preparing its first open-source LLM release, while Meta weighs opening the next LLaMA version beyond research use. A funded independent lab means the open-model layer no longer depends solely on what Meta or OpenAI choose to give away.
First-order effects
- Together now has dedicated capital to train its own open-source models, converting its RedPajama data-replication work into an independent model-building program rather than a dependency on Meta's releases.
- Buyers and developers get another non-proprietary model supplier whose roadmap doesn't hinge on Meta's licensing decisions.
Second-order effects
- A funded open-source rival sharpens the pressure behind Meta's deliberations over commercializing the next LLaMA version — releasing models freely becomes competitive defense, not just goodwill.
- OpenAI's reported open-source plan now faces a crowded field where independent labs like Together set expectations for capability, forcing incumbents to justify what they hold back.
Third-order effects
- If the pattern holds, open-source LLMs mature from a side effect of leaks into an institutionalized layer: Meta went on to ship Llama 2 free for research and commercial use and later billed Llama 3.1 as the first frontier-level open source model, validating the space Together bet on early.
- State-backed programs like the €37.4M OpenEuroLLM partnership among 20 EU organizations suggest open models are becoming infrastructure that governments fund directly, alongside venture-backed startups — a structural split between closed frontier labs and a public open-model commons.
The trend: Open-source LLM development is shifting from accidental leaks and corporate giveaways toward a deliberately funded ecosystem of startups and state-backed consortia.