Fireworks

“Not only were latency metrics nearly an order-of-magnitude less than off-the-shelf open-source options with the hardware we had access to but total cost of ownership was reasonable compared to procuring similar hardware internally.”

“Fireworks offers incredible TTFT (time-to-first-token) and ITL (inter-token latency) without the need to manage our own machines.”

“Uma is the brain behind key parts of Upwork’s hiring and matching process. Features like Uma’s ability to create tailored proposal drafts helps reduce friction for freelancers by helping them accurately communicate their services and offerings, positioning their unique skill sets and expertise to clients to stand out and win more work.”

“Fireworks has enabled the exploration and adoption of open-weight models on an enterprise scale.”

“Fireworks has been a fantastic partner in building AI dev tools at Sourcegraph. Their fast, reliable model inference lets us focus on fine-tuning, AI-powered code search, and deep code context, making Sourcegraph the best AI coding assistant. They are responsive and ship at an amazing pace.”

“The responsiveness of the Fireworks team and their fast query resolution have been remarkable. It’s been a pleasure to work with such an efficient and supportive partner.”

“Enterprise optimizations like prompt speculation and training a draft model for speculative decoding were game-changers for reducing latency.”

“The collaborative spirit of the Fireworks team was unmatched. They helped with critical inference optimizations like Flash Attention-2, n-gram speculative decoding, and benchmarking optimal replicas and shards for DS-Coder-2.”

“Fireworks’ quick support for open-source models and their ability to add newly released models rapidly allowed us to evaluate candidates offline efficiently. This significantly sped up our iteration cycles.”

““Fireworks can simplify training, but the model still learns from the examples you give it.””

““Our tool-routing and summarization projects show how fine-tuning can deliver frontier-level quality on specific, high-volume Athena tasks more efficiently. This work can improve the customer experience while lowering costs.””

““Every marketing AI team has to work out which model performs reliably on a specific task, under its own quality and operating constraints. We published AIM on the Specialized Intelligence Index so those teams can make that choice with evidence from real marketing workflows, and see where a specialized model earns its place.””

““Smaller models need clear, diverse examples that reflect the behavior we actually want them to learn. We built evaluation, data review, and synthetic data augmentation workflows so we could find gaps, improve the data, and repeat the fine-tuning process with more confidence.””

““In our experiments, fine-tuned Nemotron models from NVIDIA performed better than other tested models on this task; that is a result for our workflow, rather than a general claim about either model.””

““Fireworks provided us with infrastructure for a quick path from experimenting with open-weight models to training and serving them. Having supervised fine-tuning, preference optimization, and reinforcement fine-tuning available on one platform saves engineering time as we iterate.””

““Frontier models are excellent, and we still use them where their broader capabilities matter. But for a high-volume, repeatable domain task, a fine-tuned smaller model can deliver comparable or better quality at a lower cost per request. It can also use a shorter prompt and respond faster, which matters in an interactive product.””

“Our business model would not work without an ecosystem of inference providers with optimized inference for different models.”

“We retained full control over sampling and shaping the reward. We could focus on the RL science without having to think about kernel-level optimization, sizing, and parallelism.”

“We appreciate how quickly Fireworks validates and makes new models available for training. We know how much work goes into profiling, benchmarking, and checking numerical alignment, so that speed really matters to us.”

““The very first feedback we got from early testers of Sentient Chat was, ‘Wow, how did you get it this fast?’ It was running on Fireworks. Somehow they’re doing magic behind the scenes to make high-concurrency workloads just work.””

““Fireworks has been an exceptional partner in this mission—not only hosting RADPAIR’s advanced speech engine, but also powering our proprietary models with unmatched performance, scalability, and reliability. Their platform allows us to bring agentic intelligence in radiology to life at unprecedented speed and scale.””

““Working with RADPAIR— and through their partnership with Fireworks — has been key to pushing this forward. Now, radiologists can dictate naturally and have AI keep up in real time, which is a major leap forward,””

““The intelligence core is super important, but the gap from that core to delivering value is where the hard problem is. Talk to more users, think about product experience, do the data integrations with your customers. Offload model inference to a trustworthy provider. That’s a much better choice than managing your own hosted models, and it’s not a good use of your time when you’re building a startup.””

““We are investing in model training, fine-tuning and RL. We have the internal talent and capability to do that work, and we plan to collaborate with Fireworks on compute and training frameworks.””

““For our customers in the science community, the model-agnostic strategy is a super important factor. They don’t want model lock-in, and they want to make sure they can use state-of-the-art models.””