Chinese content西班牙推出Quasar 438B模型，欧洲最强AI
西班牙公司推出欧洲最强AI模型Quasar 438B，1M词元上下文，比Mistral Medium 3.5得分高13。
IT之家Source
Latest
Every published item, in order · 2490 items
Sep 3
Chinese content谷歌推出Gemini 3.8 Flash Cyber模型
谷歌新发布的Gemini 3.8 Flash Cyber专门用于代码安全防护，漏洞挖掘成功率超70%，修复漏洞效率是顶尖商业模型的2.6倍。
IT之家Source
Sep 2
Google Gemini 3.8 Flash model released
Available at lower cost than Gemini 3.7, this model offers better performance for software engineering tasks in the ASEAN region.
IT之家Source
World Labs launches Atlas 3D world model
Atlas offers a more efficient approach to 3D world simulation compared to specialized models, potentially reducing costs for robotics training in ASEAN.
iFLYTEK Spark X2.5 Models Released
Available for on-device deployment, these models offer longer context than alternatives like Mistral 7B for local AI applications.
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
Chinese contentGLM-5.3-Flash首秀进代码榜前十，千问3.8-Max登顶前端榜
国产模型在细分领域表现亮眼，GLM-5.3-Flash代码能力超越多数海外模型，千问3.8-Max直接登顶前端榜。
IT之家Source
Meta Muse Voice Transcribe Model Released
Muse Voice Transcribe reduces latency and failure points compared to traditional multi-system voice processing stacks.
Chinese content谷歌发布Gemini 3.8 Flash编程模型
谷歌新发布的Gemini 3.8 Flash编程能力已超过Anthropic Opus，缩小了与OpenAI的差距。
IT之家Source
Claude Fable 5.1 animated pelican test
Fable 5.1's performance on scientific benchmarks and varying reasoning levels offers developers in ASEAN a cost-effective alternative to GPT-5.6 Sol.
Chinese contentAnthropic 发布 Claude Fable 5.1 和 Mythos 5.1
Anthropic 发布两款新模型，编程和科学能力大幅提升，价格更实惠，安全防护也增强了。
IT之家Source
Chinese contentAnthropic发布Claude Fable 5.1和Mythos 5.1
Anthropic推出Claude 5.1双版本，科学基准分翻倍，缓存成本降75%，API有重大变更
Chinese contentAnthropic发布Claude Fable 5.1模型
Anthropic新发布的Fable 5.1模型在科学基准测试中翻倍得分，编码能力提升30%，同时成本降低45%。
Sep 1
AQuA framework for quantitative finance
AQuA offers a solution for financial institutions in ASEAN seeking more reliable autonomous quantitative modeling systems.
Google releases TimesFM-3 model
TimesFM-3 offers multivariate forecasting capabilities at 330 million parameters, outperforming larger models in efficiency for ASEAN businesses with time series data needs.
IT之家Source
Intelligent Chemical Industry Model 3.0 Pro Released
The model is already used by 300+ chemical enterprises in China, offering API access for complex chemical tasks with professional accuracy.
IT之家Source
WAM model achieves self-evolution
WAM offers new approach to autonomous AI learning without parameter updates, potentially reducing computational costs for developers.
Models12:58
Chinese contentFal's H3 Max Live打破视频生成速度极限Fal刚发布的H3 Max Live模型能实时生成视频，比观看速度还快，彻底改变了视频生成方式。
Tencent's Hy4 preview lightweight model released
The 214GB model enables efficient deployment on consumer hardware while maintaining performance, making advanced MoE models accessible to developers in the region.
IT之家Source
Chinese content科大讯飞开源百万 Token 端侧模型 X2.5-4B 和 1.7B
科大讯飞开源百万 Token 端侧模型，支持完整售后手册分析，可部署于机器人或边缘设备，9月7日还将发布293B版本。
IT之家Source
Aug 31
DeepSeek-V4-Flash-Vision-Exp Model Open-Sourced
Available as an experimental version with MIT License, offering multimodal capabilities close to premium models at no cost.
IT之家Source
Only the latest 2000 items are listed; for older ones see the daily archive