DeepSeek-V4-Pro刚在Fireworks上线，跑智能体和编码任务比Opus 4.8强，成本效率还高6倍，可以试试。
DeepSeek-V4-Pro-0813已在Fireworks上线，提供Day-0使用。该模型面向智能体与工具增强工作流，在Terminal Bench 2.1、Cybergym和DeepSWE基准上超越Opus 4.8，成本效率最高达6倍。它在Vals Index的Coding和Legal基准上与领先前沿模型持平。Fireworks建议Pro-0813用于重推理与编码，Flash-0731用于预算较低的任务。
DeepSeek-V4-Pro-0813 is live Day-0 on Fireworks! Built for agentic and tool-augmented workflows, it...
DeepSeek-V4-Pro-0813 is live Day-0 on Fireworks! Built for agentic and tool-augmented workflows, it outperforms Opus 4.8 on quality benchmarks like Terminal Bench 2.1, Cybergym, and DeepSWE at up to 6x the cost efficiency. It is on par with leading frontier models on Vals Index across Coding and Legal benchmarks. What to select: - Use Pro-0813 for heavy reasoning & coding - Use Flash-0731 for budget, less-intensive tasks Start building: fireworks.ai/models/deepsee… 💬 1 🔄 1 ❤️ 15 👀 1078 📊 4 ⚡