Chinese contentAI21 与 Deel 联手后训练 Nemotron 3.5 Super，3 周在薪酬领域超过 Claude Opus 5.5
AI21 拿 NVIDIA 的 Nemotron 3.5 Super 做后训练，只花 3 周就在薪酬领域跑赢了 Claude Opus 5.5。做垂直场景落地的团队可以看看他们怎么把领域数据和后训练流程结合的。
#AI21 Labs
7 stories · 1 in 7 days · 6 in 30 days
Stories, products and papers tagged "AI21 Labs", newest first.
Nov 12
Oct 22
Chinese contentAI21 Labs 在 NVIDIA GTC Berlin 分享后训练实践与客户案例结果
AI21 的 CTO 在柏林 GTC 上讲了他们对开源模型做后训练的方法，还拿 Gong、deel 等客户的实际结果做例子，想做私有化部署的可以看看。
Oct 21
Chinese contentAI21 Labs 后训练 NVIDIA Nemotron 3.5 Super，成本仅 GPT-6.1 Sol 的 1/25
AI21 后训练了 NVIDIA 的 Nemotron 3.5 Super，效果超过 GPT-6.1 Sol 但成本只有 1/25，还被 Jensen Huang 拿上主题演讲了。
Chinese contentAI21 Labs 与 Google Cloud GKE、Kueue 团队合作，生产需求回馈进 Kueue
AI21 Labs 把自己在 GKE 上跑生产任务的需求反馈给了 Kueue 团队，部分改动直接进了 Kueue，做 K8s 任务调度的可以看下博客。
Oct 20
Chinese contentAI21 Labs：用开源模型做私有化后训练，构建竞争对手拿不到的数据护城河
AI21 Labs 在 NVIDIA GTC Berlin 摆摊，讲的就是一件事：别光租 Claude 和 GPT，用开源模型加上自家数据做后训练，模型会越用越懂你的业务，竞对抄不走。
Oct 15
AI21 Labs Post-Trains NVIDIA Nemotron 3.5 Super VL for Enterprise Clients
The interesting signal is the delivery model: a lab embedding researchers with customers to fine-tune an open NVIDIA base model cheaply, instead of selling API access to a proprietary frontier model. Engineering and AI leads evaluating build-versus-buy for domain models can use the thread's results as a reference point. Deel's involvement gives it a hiring-and-HR use-case angle.
Sep 28
AI21 Labs releases two AI Engineer Fair talks online
The chunking talk signals that naive fixed-size splitting of documents is losing ground to structure-aware approaches in production RAG systems. Engineers building retrieval pipelines can use it to re-evaluate their chunking strategies, and anyone running vLLM in production gets a concrete debugging walkthrough. Watching both takes well under an hour.