Immersive translate拡張に入れると翻訳楽すぎるなこりゃ凄い🥺
ちょっと雑だから詳しく翻訳したい時はGPTやGeminiに投げるけど
Immersive translateのGLM4でもそれなりに読める🥺
こりゃ革命だ🥺
ちょっと雑だから詳しく翻訳したい時はGPTやGeminiに投げるけど
Immersive translateのGLM4でもそれなりに読める🥺
こりゃ革命だ🥺
January 11, 2026 at 1:47 AM
Immersive translate拡張に入れると翻訳楽すぎるなこりゃ凄い🥺
ちょっと雑だから詳しく翻訳したい時はGPTやGeminiに投げるけど
Immersive translateのGLM4でもそれなりに読める🥺
こりゃ革命だ🥺
ちょっと雑だから詳しく翻訳したい時はGPTやGeminiに投げるけど
Immersive translateのGLM4でもそれなりに読める🥺
こりゃ革命だ🥺
Comment "ChatGPT5" to get this NEW AI Update
🚨 GPT-5 Just Met Its Match from China!
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
🚨 GPT-5 Just Met Its Match from China!
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
October 26, 2025 at 5:40 PM
Comment "ChatGPT5" to get this NEW AI Update
🚨 GPT-5 Just Met Its Match from China!
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
🚨 GPT-5 Just Met Its Match from China!
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
Comment "ChatGPT5" to get this NEW AI Update
🚨 NEW 1-Click GPT-5 AI Agents are INSANE! 🤯
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
🚨 NEW 1-Click GPT-5 AI Agents are INSANE! 🤯
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
October 26, 2025 at 5:41 PM
Comment "ChatGPT5" to get this NEW AI Update
🚨 NEW 1-Click GPT-5 AI Agents are INSANE! 🤯
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
🚨 NEW 1-Click GPT-5 AI Agents are INSANE! 🤯
#AInews #GPT5 #ChinaAI #Qwen3 #GLM4 #TechNews #ArtificialIntelligence #JulianGoldieStyle
adaptable to complex software engineering workflows. GLM4 Flash, a large language model, leverages its advanced capabilities in natural language understanding, contextual reasoning, and multilingual support to generate [4/8 of https://arxiv.org/abs/2502.18465v1]
February 27, 2025 at 5:59 AM
adaptable to complex software engineering workflows. GLM4 Flash, a large language model, leverages its advanced capabilities in natural language understanding, contextual reasoning, and multilingual support to generate [4/8 of https://arxiv.org/abs/2502.18465v1]
Okay so there’s a bug which is causing GLM to use more memory for the kv cache than it should and this has been affecting most runtimes (vllm, llama cpp and mlx). It’s fixed upstream in all of them but I don’t think a new version with the patch has been released: github.com/ml-explore/m...
Update glm4_moe_lite To Store KV Latent in Cache by N8python · Pull Request #780 · ml-explore/mlx-lm
Updates GLM4-MOE-Lite to not store full KV in cache and instead used compressed MLA latent. Tested w/ 6-bit version of model (I don't have RAM for bf16). This works because we can move the late...
github.com
January 23, 2026 at 12:34 AM
Okay so there’s a bug which is causing GLM to use more memory for the kv cache than it should and this has been affecting most runtimes (vllm, llama cpp and mlx). It’s fixed upstream in all of them but I don’t think a new version with the patch has been released: github.com/ml-explore/m...
New video! I talk about why you must play Baten Kaitos: Eternal Wings and the Lost Ocean! Please check it out!
Link: www.youtube.com/watch?v=GLm4...
Link: www.youtube.com/watch?v=GLm4...
October 28, 2023 at 4:11 PM
New video! I talk about why you must play Baten Kaitos: Eternal Wings and the Lost Ocean! Please check it out!
Link: www.youtube.com/watch?v=GLm4...
Link: www.youtube.com/watch?v=GLm4...
Meanwhile, GLM4-Edge is now on Hugging Face hub🚀
👉 huggingface.co/collections/...
Packed with advanced dialogue + multimodal models:
📱 1.5B / 2B models: Built for mobile & in-car systems
💻 4B / 5B models: Optimized for PCs
👉 huggingface.co/collections/...
Packed with advanced dialogue + multimodal models:
📱 1.5B / 2B models: Built for mobile & in-car systems
💻 4B / 5B models: Optimized for PCs
GLM-Edge - a THUDM Collection
Unlock the magic of AI with handpicked models, awesome datasets, papers, and mind-blowing Spaces from THUDM
huggingface.co
November 29, 2024 at 8:42 AM
Meanwhile, GLM4-Edge is now on Hugging Face hub🚀
👉 huggingface.co/collections/...
Packed with advanced dialogue + multimodal models:
📱 1.5B / 2B models: Built for mobile & in-car systems
💻 4B / 5B models: Optimized for PCs
👉 huggingface.co/collections/...
Packed with advanced dialogue + multimodal models:
📱 1.5B / 2B models: Built for mobile & in-car systems
💻 4B / 5B models: Optimized for PCs
Ollama Cloud is a game-changer: access powerful open-source models without high-end hardware at a fraction of commercial API cost. This products taxonomy classification demo runs on it. Well done, Ollama! #Ollama #OllamaCloud #AffordableAI #GLM4 #OpenSourceAI #AI #LLM
matasoft.hr/QTrendContro...
matasoft.hr/QTrendContro...
An example of using (Un)Perplexed Spready in a real life use case - Furniture catalog taxonomy classification
AI-driven spreadsheet software (Un)Perplexed Spready automates complex data tasks by integrating advanced AI models directly into spreadsheets. This article demonstrates how (Un)Perplexed Spready, com...
matasoft.hr
January 10, 2026 at 2:25 PM
Ollama Cloud is a game-changer: access powerful open-source models without high-end hardware at a fraction of commercial API cost. This products taxonomy classification demo runs on it. Well done, Ollama! #Ollama #OllamaCloud #AffordableAI #GLM4 #OpenSourceAI #AI #LLM
matasoft.hr/QTrendContro...
matasoft.hr/QTrendContro...
GLM-4.7 is used for this demo—currently one of the best open-source LLMs, surpassing models like kimi-k2. See what it can do in a practical spreadsheet task of products taxonomy classification.
matasoft.hr/QTrendContro...
#GLM4 #OpenSourceAI #Benchmarks
#AIAdvantage #SpreadsheetFormula #AI #LLM
matasoft.hr/QTrendContro...
#GLM4 #OpenSourceAI #Benchmarks
#AIAdvantage #SpreadsheetFormula #AI #LLM
An example of using (Un)Perplexed Spready in a real life use case - Furniture catalog taxonomy classification
AI-driven spreadsheet software (Un)Perplexed Spready automates complex data tasks by integrating advanced AI models directly into spreadsheets. This article demonstrates how (Un)Perplexed Spready, com...
matasoft.hr
January 10, 2026 at 10:07 AM
GLM-4.7 is used for this demo—currently one of the best open-source LLMs, surpassing models like kimi-k2. See what it can do in a practical spreadsheet task of products taxonomy classification.
matasoft.hr/QTrendContro...
#GLM4 #OpenSourceAI #Benchmarks
#AIAdvantage #SpreadsheetFormula #AI #LLM
matasoft.hr/QTrendContro...
#GLM4 #OpenSourceAI #Benchmarks
#AIAdvantage #SpreadsheetFormula #AI #LLM
I mean it won’t be any different. Lm studio has an mlx backend option and it also allows for the max size of the context window to be controlled. But there’s a bug which is impacting basically all runtimes w/ this model rn.
github.com/ml-explore/m...
github.com/ml-explore/m...
Update glm4_moe_lite To Store KV Latent in Cache by N8python · Pull Request #780 · ml-explore/mlx-lm
Updates GLM4-MOE-Lite to not store full KV in cache and instead used compressed MLA latent. Tested w/ 6-bit version of model (I don't have RAM for bf16). This works because we can move the late...
github.com
January 23, 2026 at 12:37 AM
I mean it won’t be any different. Lm studio has an mlx backend option and it also allows for the max size of the context window to be controlled. But there’s a bug which is impacting basically all runtimes w/ this model rn.
github.com/ml-explore/m...
github.com/ml-explore/m...
Zhipu’s 9B citation model brings 128K context with LoRA — a credible fit for RAG pipelines chasing verifiable sources without 70B-class costs.
Qualifier first: citation scores on Chinese benchmarks rarely translate to messy English corpora, and the eval set is thin. At 9B, adversarial retrieval...
Qualifier first: citation scores on Chinese benchmarks rarely translate to messy English corpora, and the eval set is thin. At 9B, adversarial retrieval...
Zhipu AI Refines LongCite-glm4-9b for Superior Citation Accuracy
aichina.news
August 18, 2026 at 7:48 AM
Zhipu’s 9B citation model brings 128K context with LoRA — a credible fit for RAG pipelines chasing verifiable sources without 70B-class costs.
Qualifier first: citation scores on Chinese benchmarks rarely translate to messy English corpora, and the eval set is thin. At 9B, adversarial retrieval...
Qualifier first: citation scores on Chinese benchmarks rarely translate to messy English corpora, and the eval set is thin. At 9B, adversarial retrieval...
9B beating GPT-4o on citation F1 (57.5 on LongBench-Cite) is a real number, not vibes, and 128K context trained under LongAlign is what RAG pipelines have actually been waiting for.
Counterweight: LongBench-Cite is the only benchmark cited, and it's the model's own training ground. One-eval...
Counterweight: LongBench-Cite is the only benchmark cited, and it's the model's own training ground. One-eval...
Breaking News: Chinese Model zhipuai/LongCite-glm4-9b Sets New Benchmark for Citation Accuracy
aichina.news
July 23, 2026 at 6:43 PM
9B beating GPT-4o on citation F1 (57.5 on LongBench-Cite) is a real number, not vibes, and 128K context trained under LongAlign is what RAG pipelines have actually been waiting for.
Counterweight: LongBench-Cite is the only benchmark cited, and it's the model's own training ground. One-eval...
Counterweight: LongBench-Cite is the only benchmark cited, and it's the model's own training ground. One-eval...
Zhipu's belated model card for LongCite-glm4-9b lands the basics: architecture, training, citation accuracy. Useful for anyone tracing long-context output provenance.
But no independent eval, no dataset transparency. Claims stay soft without third-party replication.
Production leads evaluating...
But no independent eval, no dataset transparency. Claims stay soft without third-party replication.
Production leads evaluating...
ZhipuAI Enhances LongCite-glm4-9b with Comprehensive Documentation
aichina.news
July 22, 2026 at 4:57 PM
Zhipu's belated model card for LongCite-glm4-9b lands the basics: architecture, training, citation accuracy. Useful for anyone tracing long-context output provenance.
But no independent eval, no dataset transparency. Claims stay soft without third-party replication.
Production leads evaluating...
But no independent eval, no dataset transparency. Claims stay soft without third-party replication.
Production leads evaluating...
The model achieves 100% accuracy in the 1M length Passkey Retrieval task and scores 93.1 on the long text evaluation benchmark RULER, surpassing GPT-4’s 91.6 and GLM4-9B-1M’s 89.9.
Processing a context of 1M tokens from 4.9 minutes to 68 seconds, achieving a 4.3x speedup.
Processing a context of 1M tokens from 4.9 minutes to 68 seconds, achieving a 4.3x speedup.
November 21, 2024 at 4:44 PM
The model achieves 100% accuracy in the 1M length Passkey Retrieval task and scores 93.1 on the long text evaluation benchmark RULER, surpassing GPT-4’s 91.6 and GLM4-9B-1M’s 89.9.
Processing a context of 1M tokens from 4.9 minutes to 68 seconds, achieving a 4.3x speedup.
Processing a context of 1M tokens from 4.9 minutes to 68 seconds, achieving a 4.3x speedup.
Cloud ones,
> Grok is the best right now imo.
> Gemini is great as well.
> z.ai (glm4.x) seems alright.
> ChatGPT nah probably, because I don't like OpenAI or Sam Altman.
>Meta's AI is a waste of time, not competitive.
> Grok is the best right now imo.
> Gemini is great as well.
> z.ai (glm4.x) seems alright.
> ChatGPT nah probably, because I don't like OpenAI or Sam Altman.
>Meta's AI is a waste of time, not competitive.
February 10, 2026 at 1:29 PM
Cloud ones,
> Grok is the best right now imo.
> Gemini is great as well.
> z.ai (glm4.x) seems alright.
> ChatGPT nah probably, because I don't like OpenAI or Sam Altman.
>Meta's AI is a waste of time, not competitive.
> Grok is the best right now imo.
> Gemini is great as well.
> z.ai (glm4.x) seems alright.
> ChatGPT nah probably, because I don't like OpenAI or Sam Altman.
>Meta's AI is a waste of time, not competitive.
GLM4.7-Flash the new Local LLM king at 30B A3B and OpenCode? Article URL: https://grigio.org/glm4-7-flash-the-new-local-llm-king-at-30b-a3b/ Comments URL: https://news.ycombinator.com/item?id=46751...
Origin | Interest | Match
Origin | Interest | Match
GLM4.7-Flash the new Local LLM king at 30B A3B ?
GLM4.7-Flash represents Z.ai's latest breakthrough in the 30B parameter class, delivering a Mixture-of-Experts (MoE) model that balances high performance with efficiency. Released in January 2026, this model has quickly established itself as a formidable contender in agentic coding and general reasoning tasks. Technical Architecture Model Specifications: * Parameters: 30B
grigio.org
January 25, 2026 at 8:15 AM
GLM 4.5 vs. Promptfoo: A Playbook for Systematic LLM Security Audits GLM 4.5 by Zhipu AI marks a major leap forward in open-source LLMs, blending deep reasoning, long-context understanding, and age...
#glm4 #opensource #security #llm
Origin | Interest | Match
#glm4 #opensource #security #llm
Origin | Interest | Match
GLM 4.5 vs. Promptfoo: A Playbook for Systematic LLM Security Audits
GLM 4.5 by Zhipu AI marks a major leap forward in open-source LLMs, blending deep reasoning,...
dev.to
August 3, 2025 at 10:12 AM
发现像智谱「glm4:9b」、通义「qwen2.5:7b」、零一「yi1.5:9b」这样可本地部署且较新的国产开源聊天大模型，用来对那些又臭又长的小说进行剧情浓缩和归纳，简直是特攻。而且本地使用时不会像调用接口后触发不知道什么违禁词不给回答，更加隐私安全。
感觉这下可以速读完所有拖着进度看不完的作品，还能节约相当多的时间了。
看来人工智能真的让用上了人工智能的人们都拥有了「美好的未来」呢，写是用人工智能辅助写的，读也是用人工智能辅助读的，只有卖显卡的赚疯了的世界，终于还是要出现了吗。
感觉这下可以速读完所有拖着进度看不完的作品，还能节约相当多的时间了。
看来人工智能真的让用上了人工智能的人们都拥有了「美好的未来」呢，写是用人工智能辅助写的，读也是用人工智能辅助读的，只有卖显卡的赚疯了的世界，终于还是要出现了吗。
October 27, 2024 at 8:41 AM
发现像智谱「glm4:9b」、通义「qwen2.5:7b」、零一「yi1.5:9b」这样可本地部署且较新的国产开源聊天大模型，用来对那些又臭又长的小说进行剧情浓缩和归纳，简直是特攻。而且本地使用时不会像调用接口后触发不知道什么违禁词不给回答，更加隐私安全。
感觉这下可以速读完所有拖着进度看不完的作品，还能节约相当多的时间了。
看来人工智能真的让用上了人工智能的人们都拥有了「美好的未来」呢，写是用人工智能辅助写的，读也是用人工智能辅助读的，只有卖显卡的赚疯了的世界，终于还是要出现了吗。
感觉这下可以速读完所有拖着进度看不完的作品，还能节约相当多的时间了。
看来人工智能真的让用上了人工智能的人们都拥有了「美好的未来」呢，写是用人工智能辅助写的，读也是用人工智能辅助读的，只有卖显卡的赚疯了的世界，终于还是要出现了吗。
New on CRAN: glm4 (0.1.0). View at https://CRAN.R-project.org/package=glm4
glm4: Fitting Generalized Linear Models Using Sparse Matrices
Fits Generalised Linear Models (GLMs) with sparse and dense 'Matrix' matrices for memory efficiency. Acts as a wrapper for the glm4() function in the 'MatrixModels' package <<a href="https://doi.org/10.32614%2FCRAN.package.MatrixModels" target="_top">doi:10.32614/CRAN.package.MatrixModels</a>>, but adds convenient model methods and functions designed to mimic those associated with the glm() function from the 'stats' package.
CRAN.R-project.org
April 16, 2026 at 10:31 AM
New on CRAN: glm4 (0.1.0). View at https://CRAN.R-project.org/package=glm4
Breaking Barriers: GLM4-MoE Gets 65% Faster with SGLang Optimizations
The landscape of large language model (LLM) deployment is evolving rapidly, and Novita AI is at the forefront with a groundbreaking suite of performance enhancements for GLM4-MoE models. These models, known for their flexibility…
The landscape of large language model (LLM) deployment is evolving rapidly, and Novita AI is at the forefront with a groundbreaking suite of performance enhancements for GLM4-MoE models. These models, known for their flexibility…
Breaking Barriers: GLM4-MoE Gets 65% Faster with SGLang Optimizations
The landscape of large language model (LLM) deployment is evolving rapidly, and Novita AI is at the forefront with a groundbreaking suite of performance enhancements for GLM4-MoE models. These models, known for their flexibility and high compute demands, are now seeing dramatic reductions in latency and throughput times thanks to innovative techniques built on SGLang. From production-ready kernel optimizations to cutting-edge speculative decoding strategies, this update is a game-changer for organizations relying on agentic coding workflows.
undercodenews.com
January 23, 2026 at 12:58 PM
Breaking Barriers: GLM4-MoE Gets 65% Faster with SGLang Optimizations
The landscape of large language model (LLM) deployment is evolving rapidly, and Novita AI is at the forefront with a groundbreaking suite of performance enhancements for GLM4-MoE models. These models, known for their flexibility…
The landscape of large language model (LLM) deployment is evolving rapidly, and Novita AI is at the forefront with a groundbreaking suite of performance enhancements for GLM4-MoE models. These models, known for their flexibility…
The demo, showcasing products taxonomy classification with spreadsheet AI formula, used GLM-4.7, a model that excels at following complex instructions and tool use. Picking the right model for the job is part of the secret sauce. #ModelSelection #GLM4 #PromptEngineering
matasoft.hr/QTrendContro...
matasoft.hr/QTrendContro...
An example of using (Un)Perplexed Spready in a real life use case - Furniture catalog taxonomy classification
AI-driven spreadsheet software (Un)Perplexed Spready automates complex data tasks by integrating advanced AI models directly into spreadsheets. This article demonstrates how (Un)Perplexed Spready, combined with SearXNG, Matasoft Web Search tool and Ollama Cloud provided GLM-4.7 LLM model, can perform advanced AI-driven mapping and categorization of products to appropriate hierarchical product taxonomy.
matasoft.hr
January 24, 2026 at 2:50 PM
The demo, showcasing products taxonomy classification with spreadsheet AI formula, used GLM-4.7, a model that excels at following complex instructions and tool use. Picking the right model for the job is part of the secret sauce. #ModelSelection #GLM4 #PromptEngineering
matasoft.hr/QTrendContro...
matasoft.hr/QTrendContro...