Chinese contentDeepSWE测试显示DeepSeek V4.1-Flash输入token占比远高于输出
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。
Latest
Every published item, in order · 1031 items
Sep 16
Sep 15
Chinese contentClaude Fable 5.1 仍居 AI 指数首位，与 GPT-6 Astra 并列
Anthropic 新发布的 Claude Fable 5.1 在最新 AI 指数中表现不错，和 GPT-6 Astra 并列第一，而且能帮你省点钱。
DeepSeek-V4-Flash on AMD GPUs: Correctness and Performance Engineering
For users interested in optimizing large language model inference on AMD hardware, this paper provides detailed engineering insights and performance benchmarks.
Chinese contentDeepSeek-V4.1-Flash (Max) 发布，性能成本比提升 4.87%，成本仅 0.07 美元/任务
DeepSeek 发布了新模型，性能提升 4.87%，成本只有 0.07 美元，比很多模型便宜很多。
Chinese contentDeepSeek-V4.1-Flash (Max) 在任务成本和性能上取得新突破
DeepSeek新发布的V4.1-Flash (Max)模型，任务成本比之前低，性能还提升了4.87%，在性能和成本上都有优势。
Chinese contentDeepSeek-V4.1-Flash (Max) 在开源模型排名中位列第三
DeepSeek 新出的 V4.1-Flash (Max) 模型在 Agent Arena 排名第三，成本还特别低，适合做任务。
Sep 14
Chinese contentGPT Image 2.5 Flare 与 Recraft V4.1 Pro 的图像生成对比测试
朋友A和朋友们对比了GPT Image 2.5 Flare和Recraft V4.1 Pro生成的四张图片，你可以去看看他们用了什么提示词，看看谁画得更好。
China Mobile Cloud debuts heterogeneous LLM inference stack
This is a significant development in AI infrastructure, showing how companies are optimizing LLM performance and costs.
RoofLang: Enabling AI-Driven Architecting of LLM Inference Systems
This matters because it introduces a new tool, RoofLang, that could significantly improve the performance of LLMs like DeepSeek V4-series by optimizing their inference architecture, potentially leading to faster processing and better user experience.
Debiasing LLM-as-a-Judge evaluation with TraceJudgeBench
This paper provides a detailed analysis of how debiasing methods affect the accuracy of LLMs used as judges, which is important for anyone working with RAG or agent-workflow evaluation.
Switching from Codex to WorkBuddy: A Week's Experience
For users who rely on AI coding assistants, WorkBuddy offers a more cost-effective alternative to Codex, especially with models like GLM-5.3-flash that reduce costs by up to 90%.
Chinese content上海 AGI Bar 的 Next Token 第二期上线，讨论 GPT-6 Astra、DeepSeek V4.1 Flash 和苹果的 iPhone Duo
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。
DeepSeek routes V4-Pro API traffic to V4.1-Flash
This is a pricing change for DeepSeek's API that will affect developers using the V4-Pro model, as it will now be charged at the lower Flash rate until a new version is released.
Home AI lab setup with Qwen 3.8 and DeepSeek v4
This detailed guide is useful for anyone looking to build a local AI infrastructure, showing how to combine different models and hardware for specific tasks.
Sep 13
China's first domestic GPU + neuromorphic chip hybrid inference system for large models released
This system is a significant advancement in China's domestic AI hardware, offering a more cost-effective alternative to traditional GPU-only setups for large model inference.
IT之家Source
Sep 12
Qwen3.8-27B model runs on Cerebras for faster inference
The Qwen3.8-27B model is now available on Cerebras hardware, offering faster inference speeds compared to its previous deployment.
DeepSeek v4.1-Flash model released
This is a major upgrade for the DeepSeek model, showing significant progress in its architecture and scale.
DeepSeek V4.1 Flash benchmarked against coding agents
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.
Chinese contentDeepSeek V4.1 Flash 在 OpenRouter 上 24 小时处理 1T token
DeepSeek 新模型 V4.1 Flash 性价比很高，在 OpenRouter 上 24 小时处理 1T token，比 GLM-5.3 Flash 便宜 5 倍。
Sep 11
Chinese contentAI时代工程师技能变化：OpenAI GPT-6、Anthropic Claude Fable 5.1等新模型发布
朋友，OpenAI和Anthropic都发布了新模型，GPT-6和Claude Fable 5.1，还有Google等公司的语音转文本模型，错误率都低于4%，这些新模型和技能变化信息很实用。
DeepSeek extends V4 Pro API service
This matters to users because DeepSeek is keeping its older, more affordable model available for another month, allowing them to continue using it without a price increase.
IT之家Source
DeepSeek V4.1 Flash replaces V4 Pro
For users looking for a more affordable and efficient large language model, DeepSeek V4.1 Flash offers a clear upgrade path from the V4 Pro model.
DeepSeek-V4.1-Flash model now available on Ollama's cloud
This new model is more capable, faster, and more cost-effective than previous DeepSeek models, offering a better alternative for users seeking a powerful AI tool.
Chinese contentDeepSeek V4.1 Flash 模型发布，价格低于 OpenAI 和 Z.AI
DeepSeek 新模型比 OpenAI 和 Z.AI 便宜，性能还更好，值得看看。
DeepSeek-V4.1-Flash rolls out for Pro plan subscribers
This is a new version of the DeepSeek-V4.1-Flash model that is now available for Pro plan subscribers on Ollama's cloud.
Chinese contentDeepSeek V4.1 Flash 性能接近 GPT-5.6，成本仅为其 1/30
DeepSeek 推出了 V4.1 Flash，性能接近 GPT-5.6，但成本只有 1/30，用 15 美元就能跑 100 个复杂任务，性价比很高。
Cambricon Day-0 supports DeepSeek-V4.1-Flash
For users interested in high-performance AI chips, this shows a new hardware platform supporting a specific model, which could be relevant for those looking at alternatives to existing solutions.
PARSER agent design for long-context reasoning
PARSER's design could lead to more efficient and accurate AI assistants for processing long documents, which is relevant for any user dealing with large amounts of text.
Chinese contentDeepSeek 发布 V4.1 Flash 模型，采用不对称 MoE 架构并开源
DeepSeek 新出的 V4.1 Flash 模型，用不对称 MoE 架构，参数量 552B，但只用了 8B 活跃参数处理输入、16B 处理输出，比上一代节省了 3/4 的存储，算力效率更高，还支持多模态，价格也便宜，适合需要高效推理的场景。
SuperTechFansSource
Chinese contentDeepSeek-V4.1-Flash 在 Ollama 云端上线
DeepSeek 新出的 V4.1-Flash 模型，在 Ollama 云端上线了，比之前的小模型更智能、更快，还能处理图片，适合 Max 和 Team 帐户的用户。
Chinese contentDeepSeek 发布 V4.1 Flash 模型，全面超越 V4 Pro
DeepSeek 新发布的 V4.1 Flash 模型，推理速度更快，价格也降了 60%，比之前的 V4 Pro 强很多。
IT之家Source
Chinese contentDeepseek-V4.1-Flash 在 Fireworks 上正式发布
朋友，Deepseek 新出的 552B 大模型现在在 Fireworks 上能用，做编程和网络安全很厉害，比 GPT 5.6 Sol 便宜很多。
Products04:52
Chinese contentRecraft V4 Styles 上线 ElevenCreative 平台朋友发了 Recraft V4 Styles，能在 ElevenCreative 上用，能生成风格一致的图，比之前方便。
Chinese contentDeepseek-V4.1-Flash 模型上线 Fireworks 平台
朋友，Deepseek 新出的 552B 大模型现在在 Fireworks 上能用，编程和网络安全方面比 Opus 5 强，还便宜很多。
Chinese contentDeepSeek V4.1 Flash 模型发布，成本降低50倍，性能接近 Sol 和 Opus
DeepSeek 新出的 V4.1 Flash 模型，用起来特别便宜，比 Sol 和 Opus 便宜50倍，性能还差不多，适合预算有限的人。
Chinese contentDeepSeek 发布 V4.1 Flash 模型，性能对标 GLM-5.3，KV 缓存技术大幅降低内存占用
DeepSeek 新出的 V4.1 Flash 模型，性能不错，和 GLM-5.3 差不多，关键是内存消耗特别低，适合做 agentic 任务，价格也便宜很多。
Chinese contentClaude Fable 5.1 和 GPT-6 Astra 让专有模型与开源模型在网页开发上的差距重新扩大
Claude和GPT-6的新版本让专有模型在网页开发上又领先了，想了解最新差距的可以看看。
Chinese contentDeepSeek-V4.1-Flash 参与Agent Arena竞赛
DeepSeek新出的V4.1-Flash模型参加Agent Arena竞赛，你可以去投票看看它的表现，这个模型设计得更智能、更快、更高效。