The latest open-weight model from Mistral—Mistral Small 3—is now live on Fireworks! Fireworks is excited to partner with Mistral to be an official launch partner for the model. With Apache 2.0 licensing, blazing-fast 150 TPS generation speeds, and a 32K context window, it’s a powerful choice for builders looking for low-latency, high-efficiency AI.
Why Mistral Small 3?
Mistral Small 3 outperforms Llama 3.3 70B base on many pretraining benchmarks while being 3x faster on the same hardware. As the most knowledge-dense model in its class, it’s an excellent choice for:
✅ Conversational AI – Quick, accurate chatbot responses
✅ Function calling & automation – Low-latency execution for agentic workflows
✅ Fine-tuning & domain expertise – Ideal for specialized knowledge (legal, healthcare, finance)
✅ Local inference – Runs on an RTX 4090 or MacBook with 32GB RAM
Small Models, Big Models, and Compound AI Systems
At Fireworks, we believe the future of AI isn’t one monolithic model—it’s about building intelligent systems by combining specialized models. Small models like Mistral Small 3 and large models like DeepSeek V3 or GPT-4o play complementary roles in AI architectures:
🔹 Small models (like Mistral Small 3) are optimized for speed, cost, and efficiency—handling 80% of everyday tasks with ultra-low latency. These are perfect for fast-response chatbots, function calling, and local inference.
🔹 Big models excel at deep reasoning, planning, and complex problem-solving—but they come with higher computational costs and latency.
🔹 Compound AI systems use small models for routine tasks and delegate complex reasoning to larger models when needed. This hybrid approach gives developers better performance, lower costs, and more flexibility in building real-world applications.
Mistral Small 3 on Fireworks: Try It Today
Mistral Small 3 is now available both serverless and on-demand on Fireworks, with instant API access for easy experimentation and deployment. Whether you’re optimizing for speed, cost, or accuracy, Fireworks makes it easy to test and integrate models into compound AI workflows.
