LMArena测了Jev Router的实际延迟:中位数6.18秒,比DeepSeek V4.1 Flash的3.64秒慢了一截,P90差距更明显。
Latest
Routing adds a decision layer that can match the Pareto frontier yet still lose to simply picking one strong model, because latency and overhead eat the gains. Builders weighing a router against a default model should benchmark the direct path first, as the comparison here shows. The steerability signal is also useful for teams whose workflows depend on following complex instructions.
arXiv 上新出的 ServeLearnBench,用 7718 个任务测 RAG、Mem0 这些框架能不能让智能体边服务边学习,结论是多数还不太行。
DeepSeek 的 V4.1 Flash 跑分出来了,ARC-AGI-2 拿下 72.9%,比上一代高了 11.5 分,不过每任务成本贵了两倍多,看你怎么权衡。
Mistral放出Mistral Large 4,1T参数只激活49B,跑分被DeepSeek V4.1 Flash压了一头,10月就开源,可以先试试API。
Reflection AI 开源了 Beam,501B 参数 MoE,作者称推理算力只要 GLM-5.2 的三分之一,Apache 2.0 可商用,不过注意这数字只是估算。
Fireworks 现在能直接微调 DeepSeek V4.1 Flash 了,想做编程智能体或工具调用的可以看看,托管训练省不少事。
DeepSeek V4.1 Flash 把和 Anthropic 的跑分差距从 15% 追到只剩 3%,想知道中美模型现在差多少的看这篇。
一条关于 DeepSeek V4.1 的训练观察:没做 dense warmup 也能训练顺利,作者解释了为什么这种设计值得 scale up。
DeepSeek V4.1 Flash 的论文在 HF 榜单只有 191 赞,发帖人直接开怼,评论区吵起来了,看看大家怎么说。
The reveal of DSec provides insights into how DeepSeek optimizes its massive model training infrastructure, which could influence industry approaches to resource allocation. For developers and researchers working with large models, understanding this distributed architecture could inform their own training strategies and resource planning.
一个人开发者晒账单:RSS 翻译全跑 DeepSeek V4.1 Flash,每天不到 1 块钱,想给工具挑便宜模型的可以参考。
用 opencode 的 10 美元套餐算了笔账:MiMo 能跑 3 万次请求,比 DeepSeek 的 2.6 万还多,就是不太稳。想省钱写代码的可以看看怎么选。
在 VS Code 里装个 Cline 插件,Gemini 3.8 Flash、DeepSeek V4.1 Flash 这些模型都能白嫖,写代码零成本,试试看。
Permanently quadrupling the credit signals that DeepSeek's flash-tier economics are now viable for third-party platforms to subsidise, a shift in how cheap frontier-adjacent inference has become. Developers comparing coding-assistant plans can treat this as a direct price benchmark against comparable Claude or GPT flash-tier offerings. DeepSeek's official partnership with OpenCode also shows Chinese labs distributing through agentic coding tools rather than APIs alone.
OpenCode 把给 DeepSeek v4.1 Flash 的 60 美元用量额度改成永久了,用这个工具薅便宜算力的可以放心续命。
几个新模型在治理审查门上分数差距很大,完整路线跑通率最高也只有七成六,作者把为什么开源了,做流程自动化的可以看看。
Stanford 联合 Terminal-Bench 团队搞了个科研智能体榜单,70 个真实科研任务里 GPT-6 Astra 拿 63% 排第一,开源模型才 10%,差距很直观。
Cline 里免费用 Gemini 3.8 Flash,1M 上下文加 291 tok/s,跑分还压过 GPT-5.6 Luna 和 DeepSeek v4.1 Flash,写代码可以直接试。
有人拿 DeepSeek V4.1 Flash 和 Space Bunny Alpha 现场比拼,让两个模型各自写一个网页版 Minecraft,前者能跑能挖,后者地面还有破洞,结果一目了然。
Zenith 用最便宜的 DeepSeek v4.1 Flash 打赢了 GPT 5.6 Sol,开源党可以看看它怎么做到的。
有人推测 DeepSeek V4.1 会用 YOCO 这类新架构,如果靠谱,KV 缓存开销可能大幅下降,架构党可以留意。
Bolt 公了一周真实用量:GLM 5.3 Flash 占 63%,DeepSeek Flash 17%,用户宁可要快的也不挑最大的,这个数据挺说明问题。
The compute gap is smaller than it looks, because efficiency gains let Chinese labs deliver competitive inference at a fraction of the infrastructure cost. Developers and procurement teams choosing between US and Chinese models should track OpenRouter usage rankings and token pricing rather than headline capex figures. Goldman expects the next competitive front to be virtual-worker use cases in finance, legal and accounting workloads.
小米的 MiMo v2.6 Flash 跑分接近 Grok 4.7,成本却只有 1/57,做 agent 的可以看看这份 60 任务实测。
Grok 4.7 浏览器任务 39.9 分,比上代涨 8.7,但只有 GPT-6 Astra 一半。
OpenAI 用 1 万个代理做了个实验,结果挺有意思的,DeepSeek 也出了个新版本,对长文本处理更便宜了。
智谱AI新出的GLM 5.3 FlashX模型,性能确实有提升,但价格比之前贵不少,而且被DeepSeek的v4.1 Flash模型给超越了。
Bolt 上开源模型首周,开发者选最快 DeepSeek V4.1 Flash,该版本在 Bolt Forge 中上线,使用量是 DeepSeek V4 Pro 的 10 倍。
Notion 推出了 DeepSeek V4.1 Flash 模型,你可以用它来创建更快的智能体,比之前版本更快。
This shows how AI can optimize game performance on older hardware, potentially making more games playable on mid-range devices.
For those interested in the future of AI development, Liu Shengyu's perspective offers a rare and candid look at the challenges and motivations of a leading engineer working on a major model.
DeepSWE测试发现DeepSeek V4.1-Flash的输入token消耗远高于输出,缓存命中率高,能大幅降低成本。
朋友录的播客,周一摸鱼时可以听,内容有 GPT-6、DeepSeek V4.1 和苹果的 iPhone Duo 这些热点。
This is a major upgrade for the DeepSeek model, showing significant progress in its architecture and scale.
For developers, this matters because it shows DeepSeek V4.1 Flash's coding capabilities are robust and not tied to a specific agent framework, making it a versatile choice for different coding tasks.