Fireworks 推出了其训练API和Fireworks Lab，为训练和部署自定义AI模型提供专业基础设施。训练API提供Serverless和Dedicated两种计算模式，以满足不同速度和控制需求。Fireworks Lab允许公司嵌入Fireworks的研究人员和工程师共同设计或构建模型，早期采用者包括Cursor、Harvey和Figma。 AI
影响 为公司训练和部署自定义AI模型提供专业基础设施，可能加速细分领域的AI发展。
排序理由 Fireworks是一家提供推理基础设施的公司，正在推出新服务，而不是发布新基础模型的前沿实验室。
在 X — Fireworks (inference infra) 阅读 →
AI 生成摘要 · Google Gemini · 来自 7 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
Today, we also open the door to Fireworks Lab: Embed forward-deployed researchers and engineers to co-design or build your model for you. Join leaders like @cursor_ai, @harvey, @figma, @vercel, @cognition, @factoryai, and @tryheidi beating the frontier with Fireworks. Learn
→ rollout throughput: generation speed and GPU utilization → weight sync: fresh weights on the rollout deployment every step → correctness: trainer and inference engine agreeing on the same model Cofounder @jamesr66a on how we built Fireworks for all three:
Train, sample, serve, retrain on one platform, built for reinforcement learning (RL) at scale. RL couples training and inference into one loop: sample rollouts, score them, update the weights, resample. Doing that well at scale comes down to three things.
Our Training API offers two compute modes: Serverless when speed to experiment matters most, or Dedicated when scale, control, and GPU economics take priority. You focus on refining your learning signal, not building infra.
Connect your custom training loop to Fireworks-managed distributed training and rollout infrastructure. Orchestrate the loop in Python wherever you choose, with full control over your loss, reward, data, and environment.
Working with ML teams around the world surfaced where training workflows break down: constrained model and method choice, limited loop control, idle and costly compute, and split training and serving infrastructure that destabilizes runs Those partnerships shaped our Training
The most ambitious companies are training models to outperform the frontier on the capabilities that differentiate their business. Today, we announce the general availability of our Training API and Fireworks Lab, making model specialization accessible to all organizations. http…