# Own your specialized intelligence > Fireworks’ state of the art training and inference platform take you beyond the frontier, transforming open models into your specialized intelligence. ## Intelligence is power. Own yours. The specialized intelligence platform One platform to run, ensemble, and train open models with world-class researchers and uncompromising quality. [Get started](https://app.fireworks.ai/signup) · [Talk to experts](https://fireworks.ai/contact) Logos: Samsung, Uber, MongoDB, Notion, Genspark, Microsoft, Doordash, Vercel, HubSpot, Cursor, Lovable, Unity, Mercado Libre, Zillow ## Own your models., data., AI economics. Own your future. Build intelligence that sets you apart. Start with state-of-the-art open models. Ensemble them to lower costs. Train on what makes your business unique. - **Diversify your models**: Unshackle from closed frontier pricing and run open model inference for workloads where they meet the performance bar. - **Ensemble your models**: Further improve your AI economics by intelligently routing models at each turn, optimizing for accuracy and cost per task. - **Train your models**: Build a durable competitive advantage by leading the frontier on your domain. Protect your AI and data sovereignty. ## Sign up to experience Forge on demand Fireworks Forge 2026 Access the keynote now. More Forge sessions coming soon. [Access now](https://fireworks.ai/forge-2026) ## The specialized intelligence platform trusted by 10,000+ companies globally [Explore the platform](https://fireworks.ai/product) ### Inference Exceptional inference for the multi-model organization. Optimized for speed, quality, and cost from kernels to cache. Your cloud or ours. - Run the latest open models and your own at up to 4x the throughput of providers wrapping vLLM. - Get lab-designed quality. Every model is rigorously tested against the lab’s reference (KLD <0.007). - Pay less per finished task. Achieve up to 99.6% cache hits. Cached input costs half or less. - Scale globally. Keep data local. One endpoint, 30+ regions, with policy-aligned data residency. Start serverless; reserve throughput as you grow. [Explore inference](https://fireworks.ai/product/inference) ### Model ensembling Deploy the best combination of models behind your team and products. Pay frontier prices only where the task earns it. **Fireworks Nexus, for code and co-work** - Save 50% or more on AI. Route routine work to open models and reserve closed models for the hardest tasks. - Keep your tools. Nexus works with leading harnesses. Set spend limits and track savings. **Fireworks Platform, for agents and apps** - Run and configure models. Build through API, SDK or CLI, with tool calling and structured outputs to fit your application. - Train your way. Zero infrastructure to manage. Move from serverless to dedicated training with the same Training API. [Explore Nexus](https://fireworks.ai/product/nexus) · [Visit Nexus docs](https://docs.fireworks.ai/nexus) ### Training Build the model only you can on the world’s leading training platform. Any method. Deep control. Near-zero train-serve KL divergence. - Outperform general models on your tasks. Train on your data. Measure every run against your benchmarks. - Train your way. Run managed SFT, DPO and RL, or build custom loops. Scale from LoRA to full-parameter on the largest MoE models. - Deploy with training fidelity. Shared numerics, kernels and MoE expert routing preserve checkpoint behavior in production. Deploy in one click. - Iterate faster on the same budget. Train, deploy, evaluate and retrain on one platform. No GPUs to procure. Teams report 2–4x faster iterations. [Explore training](https://fireworks.ai/product/training) ### Fireworks Lab World-class researchers with reinforcement learning experience across 10,000+ GPUs. - Start with a clear path to value. Assess capabilities, baselines and data readiness. Define success. - Hill climb the metrics your business runs on. Co-design evals with your domain experts, then train to exceed performance targets. - Get to market faster. With deep experience, we’ll tailor your training strategy across methods, orchestration and compute for speed and quality. - Own everything we build. Your model, eval harness, data pipelines, training loop and recipes. [Talk to experts](https://fireworks.ai/contact-training) ## Your specialized intelligence lives here Build a model system that’s uniquely yours and operates at peak performance through co-optimized inference, ensembling, and training. [Product overview](https://fireworks.ai/product) ## Run the latest open models as leading labs designed them. Each model is custom deployed and benchmarked for optimized cost, speed, and quality. Which will you try first? [Browse models](https://fireworks.ai/models) - **[Kimi K3](https://fireworks.ai/models/fireworks/kimi-k3)**: Moonshot AI · $3/M Input · $15/M Output · 1,048,576 Context · Vision - **[DeepSeek V4.1 Flash 1017](https://fireworks.ai/models/deepseek-ai/deepseek-v4p1-flash-1017)**: New · Deepseek · $0.3/M Input · $1.2/M Output · 1,048,576 Context · Vision - **[MiniMax M3.1 Flash](https://fireworks.ai/models/fireworks/minimax-m3p1-flash)**: New · MiniMax · $0.15/M Input · $0.6/M Output · 1,048,576 Context · Vision - **[Muse Spark Open](https://fireworks.ai/models/fireworks/muse-spark-open)**: New · Meta · $1.25/M Input · $4.25/M Output · 1,048,576 Context · Vision - **[GLM-5.3](https://fireworks.ai/models/fireworks/glm-5p3)**: Z.ai · $1.4/M Input · $4.4/M Output · 1,048,576 Context · LLM - **[GLM 5.3 Flash](https://fireworks.ai/models/fireworks/glm-5p3-flash)**: Z.ai · $0.15/M Input · $0.5/M Output · 1,048,576 Context · Vision - **[Qwen 3.8 Max](https://fireworks.ai/models/fireworks/qwen3p8-max)**: Qwen · $2/M Input · $6/M Output · Vision - **[OpenAI gpt-oss-20b](https://fireworks.ai/models/fireworks/gpt-oss-20b)**: OpenAI · 131,072 Context · LLM - **[NVIDIA Nemotron 3 Ultra NVFP4](https://fireworks.ai/models/fireworks/nemotron-3-ultra-nvfp4)**: NVIDIA · 262,144 Context · LLM ## What our customers are saying > "Fireworks enabled us to **own our AI journey**, and unlock better quality in just four weeks." > > — Kay Zhu, CTO | Genspark [Read customer story](https://fireworks.ai/customers/genspark-gen-1-slides-rl-post-training) > "Fireworks has been a key partner in helping us train and serve the models behind Cursor at scale. Their platform supports the high-throughput RL workloads and production inference required for Composer, giving us the speed, reliability, and efficiency to keep pushing the frontier of AI coding." > > — Sualeh Asif, CPO | Cursor [Read customer story](https://fireworks.ai/customers/cursor-composer-2-reinforcement-learning) > "Vercel’s v0 model is a composite model. The SOTA in this space changes every day, so you don’t want to tie yourself to a single model. Using a fine-tuned reinforcement learning model with Fireworks, we perform substantially better than SOTA." > > — Malte Ubl, CTO | Vercel > “Open-weight models are going to be the main way engineering teams keep automating more while managing cost in a sustainable way. Fireworks is what enables us to deliver on that mission.” > > — Leo Tchourakov, Member of Technical Staff | Factory AI [Read customer story](https://fireworks.ai/customers/factory-open-model-case-study) ## Build your own frontier. Start running open models in minutes. Then ensemble, train and climb toward intelligence no one else can rent. [Get started](https://app.fireworks.ai/signup) · [Talk to experts](https://fireworks.ai/contact)