According to a recent LinkedIn post from Together AI, the company’s research group is exploring how large language models can improve database query optimization. The post describes a system, DBPlanBench, that exposes a database’s physical operator graph to an LLM, which then applies targeted JSON patch edits to adjust join ordering without regenerating the full plan.
Claim 30% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
The company’s LinkedIn post highlights benchmark results on TPC‑H and TPC‑DS workloads, suggesting up to 4.78x speedups on complex multi‑join queries and more than 5% improvement on 60.8% of tested queries. The content also notes substantial memory savings on specific queries and emphasizes that both the paper and code are open source, indicating a strategy focused on community adoption and technical credibility.
For investors, the post suggests Together AI is positioning itself at the intersection of generative AI and data infrastructure, a segment drawing strong enterprise interest and budgets. Demonstrated performance gains in query execution could make the company’s technology attractive for cost‑conscious data platforms and analytics users, potentially supporting customer acquisition and higher-value enterprise contracts.
The emphasis on open-source release may help accelerate experimentation and integration by developers, which can expand the ecosystem around Together AI’s tools. If the approach proves robust in production environments beyond benchmarks, it could strengthen the company’s differentiation versus other AI infrastructure providers and enhance its long-term competitive positioning in the database and AI tooling markets.