Chinese contentClaude Pro Opus 5.5与DeepSeek V4.1 Flash每 token成本相同
Claude和DeepSeek两大模型订阅价格对比，帮你算算哪个更划算
Latest
Every published item, in order · 333 items
Oct 9
Chinese contentAgenticBBO-Bench：评测 LLM 智能体黑盒优化能力的跨域基准
想用 LLM 智能体做黑盒优化的可以看看，五个领域统一评测，还拆了哪些设计选择真有用、哪些是伪收益。
Chinese contentMine Odyssey 基准用 Minecraft 重建真实地点评测空间智能体
拿 Minecraft 重建曼哈顿、白金汉宫这些真实地点来考模型找路，GPT-6 Astra 考了 85.6 分，开源的 DeepSeek-V4.1-Flash 才 23.9，差距很直观。
ByteDance Seed finds phase sensitivity behind DeepSeek long-context drift
Benchmark averages can mask periodic failure modes in long-context retrieval, so a model that looks solid on mean scores may still drop retrievals depending on where content sits in the compression window. Engineers running DeepSeek or similar KV-compressed models on RAG or agent pipelines should test retrieval across phase positions rather than relying on aggregate accuracy. The arXiv paper (2609.36322) offers the evaluation methodology to replicate.
DeepSeek V4.1 Flash n-gram editing
This approach offers a lightweight way to update models without full retraining, potentially reducing costs for personalized AI applications. Developers working with large language models could implement this technique for efficient knowledge updates, especially useful for applications requiring frequent factual updates or user-specific customization.
Oct 8
Chinese contentAnthropic Haiku 5.5 降价被指瞄准国产模型定价
博主对比了 Haiku 5.5 和 DeepSeek-V4.1 flash、GLM 5.3 flash 的价格和 Terminal Bench 4.0 分数，顺便吐槽了评论区骂战，看看各方实际差距。
vLLM v0.31.0 release
This release significantly improves inference efficiency through advanced attention mechanisms and memory management, enabling faster processing of large models. Developers working with LLMs in production environments will benefit from the optimized serving capabilities and experimental features like engine snapshots, particularly for models requiring high-throughput processing.
Chinese content韩国多家银行遭网络攻击，CrowdStrike 称疑似一人借助 AI 工具完成
一个人用 ARTEX 加上 DeepSeek、GLM、Grok 和 Claude Code 就能打穿韩国多家银行，CrowdStrike 的报告把工具链讲得很细。
Anthropic's Haiku 5.5 undercuts DeepSeek and GLM on price
The takeaway is that Anthropic is now competing on price, not just capability, in the small-model tier where Chinese labs have dominated. Builders choosing a cheap workhorse model for high-volume tasks can add Haiku 5.5 to a shortlist alongside DeepSeek-V4.1 flash and GLM 5.3 flash and re-run their own workloads to compare latency and cost. Pricing and availability for ASEAN API users should be checked against local provider rates.
vLLM Highlights DeepSeek-V4.1-Flash SWA Optimization
SWA bounded replay shows how KV-cache reuse for agent loops can be engineered around sliding-window attention, a pattern many teams will copy as agent traffic grows. Engineers running agent stacks on vLLM can apply it to reduce TTFT under high concurrency. The writeup also signals that DeepSeek-V4.1-Flash inference performance is being actively tuned by the community rather than relying on the lab alone.
Agent Arena tests show Jev Router trails DeepSeek V4.1 Flash on cost
The evaluation suggests routers are judged less on raw cost-performance and more on how quickly they adapt to user feedback, where Jev Router nearly matches Claude Opus 5.5 (High). Teams building multi-model agent pipelines can use these latency and cost figures to decide whether a routing layer beats picking DeepSeek V4.1 Flash directly. The full thread contains additional session-level insights.
Chinese contentJev Router 中位延迟 6.18 秒，高于 DeepSeek V4.1 Flash 的 3.64 秒
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。
Jev Router Trails Direct DeepSeek V4.1 Flash on Cost
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.
Chinese contentvLLM 三周优化 DeepSeek-V4.1-Flash：低并发提速 1.9 倍
vLLM 花三周把 DeepSeek-V4.1-Flash 跑快了 1.9 倍，AgentX 上吞吐 5.3 倍，还附交互图表拆解过程，做部署的可以看看。
Oct 7
Chinese contentServeLearnBench 发布：评测智能体能否从服务经验中自我改进
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。
Oct 6
Chinese contentDeepSeek V4.1 Flash 在 ARC-AGI 验证集上超越 V4 Flash
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。
Chinese contentMistral发布Mistral Large 4：1T总参数49B激活，10月开源
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。
Chinese contentReflection AI 开源 Beam：501B 参数 MoE，号称最强西方开源模型
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。
Chinese contentFireworks AI 平台支持 DeepSeek V4.1 Flash 微调训练
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。
Oct 5
Chinese contentDeepSeek V4.1 Flash 发布后，中美顶尖模型 LiveBench 差距缩至 3%
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。
Chinese contentCline 因滥用问题暂停 DeepSeek-V4.1-Flash 免费推广
Cline 暂停了 DeepSeek-V4.1-Flash 免费额度，说是被滥用太狠了，用这家的朋友可以看下还能不能白嫖。
Oct 4
Chinese content观点：DeepSeek V4.1 训练复杂度更高但信息流更简洁
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。
Chinese contentDeepSeek V4.1 Flash 论文在 HF 9 月榜单仅获 191 赞引争议
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。
Oct 3
Chinese contentGeneralityLabs Exploit Bench测试结果
GeneralityLabs公布Exploit Bench测试结果，V4.1 Flash性能下降，Claude Code超越Codex。
GLM-5.3 matches Claude in vulnerability testing
The performance gap between open-source and closed models is narrowing faster than expected, with GLM-5.3 demonstrating competitive vulnerability exploitation capabilities. Developers and security researchers should evaluate this model for ethical hacking and security testing applications, especially given its open weights nature that allows for customization and inspection.
Oct 2
Oct 1
Chinese contentQwen3.8-Max 修复 PVE 服务器故障
Qwen3.8-Max 实际修复服务器案例，对比 Grok-4.7 和 DeepSeek-V4.1-Flash 表现更优。
Fyan Human-AI Harness for Math Formalization
FYAN introduces evidence-grounded semantic auditing that ensures formal statements preserve informal specifications through explicit proof-transfer obligations. Researchers working on formal verification and theorem proving can use this approach to improve accuracy in mathematical formalization, while developers can leverage ODENumLib, the 9,355-line Lean library for numerical analysis of ordinary differential equations that FYAN built.
Cross-World On-Policy Distillation for VLMs
This research addresses a critical gap in VLM distillation where students can match teacher outputs without relying on the same visual evidence. The CW-OPD method provides a corrective signal that makes visual evidence reliance explicit rather than implicit. Developers working with smaller models can leverage this approach to achieve performance comparable to much larger models while maintaining better visual grounding.
Sep 30
DeepSeek unveils DSec training platform
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.