Log in
Research report
华福证券-计算机行业定期报告：开源模型重磅出击，设备更新&消费品以旧换新政策加码-240728
》Lama3.1系列模型表现出色 Llama3.1是由Meta公司发布的开源模型系列，最大参数量达到4050亿。它在多个benchmark 测试中表现出色，与gpt 4o和claude 3.5sonnet等闭源 SOTA模型性能相当，甚至在某些方面实现了超越。Llama3.1系列模型的上下文长度扩展到128K，支持八种语言，并且在训练过程中，Meta公司使用了超过15万亿个token，优化了训练堆
Read full report →Related industries
No related institutions
Similar reports
- Environmental stewardship for leaner AI agent systems: The case for architectural restraintArchitectural restraint can help AI agents reduce unnecessary computation, use computing resources more efficiently, and strengthen governance.Article•9-min read
- What Does Venture Capital Actually Finance? Scale and Structure around the World
- You are not a model. Don’t price per token.
- 2025 State of AI Report: The Builder’s Playbook
- AI Adoption Index
- Cursor + SpaceXAI: the fastest iterating team wins