deepseek-v4.1IMAGE
DeepSeek V4.1 Flash 1017: coding and agent gains, still behind Opus 5 on reasoning
Versus the previous version, coding, terminal use, agent execution and complex tasks improved notably, beating GPT-5.6-Sol on several specialized benchmarks, though Claude Opus 5 still leads on general reasoning and hard coding.
Geek
@geekbbArchived 10 days agoOriginally posted 2026.10.17
Geek @geekbbOriginally posted 2026.10.17
DeepSeek V4.1-Flash-1017 相比上一版提升明显，尤其在编程、终端操作、Agent 执行和复杂任务方面。 它已经能在多个专项基准中超过 GPT-5.6-Sol，但在综合推理和部分高难度编程任务上，Claude Opus 5 仍然领先。 https://api-docs.deepseek.com/zh-cn/updates/
x.com/geekbb/status/2111382679954592037Archived 2026.10.18 16:48