The updated DeepSeek-R1-0528 model delivers major improvements in reasoning, inference, and accuracy through enhanced post-training optimization and greater computational resources. It now performs at a level approaching top-tier models like O3 and Gemini 2.5 Pro, with notable gains in complex tasks such as math and programming. In the AIME 2025 benchmark, accuracy jumped from 70% to 87.5%, supported by deeper reasoning (23K vs. 12K tokens per question). The update also reduces hallucinations, improves function calling, and enhances the coding experience.
Deepseek R1 05/28 is available via Fireworks' serverless API, where you pay per token. There are several ways to call the Fireworks API, including Fireworks' Python client, the REST API, or OpenAI's Python client.
Learn MoreDeepseek R1 05/28 can be fine-tuned on your data to create a model with better response quality. Fireworks uses low-rank adaptation (LoRA) to train a model that can be served efficiently at inference time.
Learn MoreOn-demand deployments allow you to use Deepseek R1 05/28 on dedicated GPUs with Fireworks' high-performance serving stack with high reliability and no rate limits.
Learn MoreDeepseek
LLM
160K
Available
Available
$3 / $8
