GLM-5 also dropped
February 11, 2026 at 12:40 PM
GLM-5 also dropped
“pony-alpha” on openrouter is GLM-5
2x bigger than GLM-4.6
2x bigger than GLM-4.6
February 10, 2026 at 2:58 AM
“pony-alpha” on openrouter is GLM-5
2x bigger than GLM-4.6
2x bigger than GLM-4.6
... plus I got the new open weights GLM-5 1.5TB (!) model to draw me a pelican riding a bicycle simonwillison.net/2026/Feb/11/...
GLM-5: From Vibe Coding to Agentic Engineering
This is a huge new MIT-licensed model: 754B parameters and 1.51TB on Hugging Face twice the size of GLM-4.7 which was 368B and 717GB (4.5 and 4.6 were around that …
simonwillison.net
February 11, 2026 at 7:21 PM
... plus I got the new open weights GLM-5 1.5TB (!) model to draw me a pelican riding a bicycle simonwillison.net/2026/Feb/11/...
Babe It's 2026. We have GLM Flash, Qwen Coder next, Step 3.5 and Iquestcoder
I dream of having Sonnet 4.5 capabilities on a local machine
February 11, 2026 at 9:26 AM
Babe It's 2026. We have GLM Flash, Qwen Coder next, Step 3.5 and Iquestcoder
You can now train MoE models 12× faster with 35% less VRAM via our new Triton kernels (no accuracy loss).
Train gpt-oss locally on 12.8GB VRAM.
In collab with @hf.co, Unsloth trains DeepSeek, Qwen3, GLM faster.
Repo: github.com/unslothai/un...
Blog: unsloth.ai/docs/new/fas...
Train gpt-oss locally on 12.8GB VRAM.
In collab with @hf.co, Unsloth trains DeepSeek, Qwen3, GLM faster.
Repo: github.com/unslothai/un...
Blog: unsloth.ai/docs/new/fas...
February 10, 2026 at 3:37 PM
You can now train MoE models 12× faster with 35% less VRAM via our new Triton kernels (no accuracy loss).
Train gpt-oss locally on 12.8GB VRAM.
In collab with @hf.co, Unsloth trains DeepSeek, Qwen3, GLM faster.
Repo: github.com/unslothai/un...
Blog: unsloth.ai/docs/new/fas...
Train gpt-oss locally on 12.8GB VRAM.
In collab with @hf.co, Unsloth trains DeepSeek, Qwen3, GLM faster.
Repo: github.com/unslothai/un...
Blog: unsloth.ai/docs/new/fas...
been experimenting with tiny LLMs lately (like GLM 2.5 1.2B) and as a "im offline and don't have a search engine but i need to look something up" thing they're pretty ok
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
February 10, 2026 at 12:58 PM
been experimenting with tiny LLMs lately (like GLM 2.5 1.2B) and as a "im offline and don't have a search engine but i need to look something up" thing they're pretty ok
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
early vibes: glm-5 feels a little more scattershot than opus 4.5/4.6. like, it forgets what directory it's in a little more. also, I assume they're just getting pummeled with traffic, but it's a bit slow right now.
verdict: "they're charging you HOW little for this?"
verdict: "they're charging you HOW little for this?"
February 12, 2026 at 1:23 AM
early vibes: glm-5 feels a little more scattershot than opus 4.5/4.6. like, it forgets what directory it's in a little more. also, I assume they're just getting pummeled with traffic, but it's a bit slow right now.
verdict: "they're charging you HOW little for this?"
verdict: "they're charging you HOW little for this?"
While Seedance 2.0’s videos are all over the timeline, DeepSeek quietly pushed a new model update in its app.
GLM-5, Ming-flash-omni from Ant Group , MiniCPM-SALA from OpenBMB, and the upcoming MiniMax M2.5 keep the heat on 🔥
Spring Festival is around the corner, no one’s sleeping!
GLM-5, Ming-flash-omni from Ant Group , MiniCPM-SALA from OpenBMB, and the upcoming MiniMax M2.5 keep the heat on 🔥
Spring Festival is around the corner, no one’s sleeping!
2026 February 🏮 - China Open Source Highlights - a zh-ai-community Collection
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
February 11, 2026 at 9:29 PM
While Seedance 2.0’s videos are all over the timeline, DeepSeek quietly pushed a new model update in its app.
GLM-5, Ming-flash-omni from Ant Group , MiniCPM-SALA from OpenBMB, and the upcoming MiniMax M2.5 keep the heat on 🔥
Spring Festival is around the corner, no one’s sleeping!
GLM-5, Ming-flash-omni from Ant Group , MiniCPM-SALA from OpenBMB, and the upcoming MiniMax M2.5 keep the heat on 🔥
Spring Festival is around the corner, no one’s sleeping!
Z AI released GLM-5
It is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.
It is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.
February 11, 2026 at 5:49 PM
Z AI released GLM-5
It is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.
It is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.
Chinese AI startup Zhipu releases new flagship model GLM-5 reut.rs/4qyP23Y
Chinese AI startup Zhipu releases new flagship model GLM-5
China's Zhipu AI released its latest artificial intelligence model on Wednesday, joining a wave of domestic rivals unveiling more sophisticated versions of the technology ahead of the Lunar New Year festival as competition heats up in the sector.
reut.rs
February 11, 2026 at 5:30 PM
Chinese AI startup Zhipu releases new flagship model GLM-5 reut.rs/4qyP23Y
looking forward to the GLM 5 report tbh
February 11, 2026 at 7:54 PM
looking forward to the GLM 5 report tbh
GLM-5: From Vibe Coding to Agentic Engineering
https://z.ai/blog/glm-5
https://news.ycombinator.com/item?id=46977210
https://z.ai/blog/glm-5
https://news.ycombinator.com/item?id=46977210
February 11, 2026 at 8:14 PM
GLM-5: From Vibe Coding to Agentic Engineering
https://z.ai/blog/glm-5
https://news.ycombinator.com/item?id=46977210
https://z.ai/blog/glm-5
https://news.ycombinator.com/item?id=46977210
Halluzinationsrate 30 %. Das beste Modell❗️
Und es gibt immer noch "Experten" die glauben, auf Basis aktueller KI-Systeme (LLMs) sei eine Superintelligenz (AGI) erreichbar.
LLMs werden immer halluzinieren. Das ist systemimmanent.
LLMs sind leistungsfähige sprachliche Heuristikmaschinen. Mehr nicht.
Und es gibt immer noch "Experten" die glauben, auf Basis aktueller KI-Systeme (LLMs) sei eine Superintelligenz (AGI) erreichbar.
LLMs werden immer halluzinieren. Das ist systemimmanent.
LLMs sind leistungsfähige sprachliche Heuristikmaschinen. Mehr nicht.
February 10, 2026 at 5:27 AM
Halluzinationsrate 30 %. Das beste Modell❗️
Und es gibt immer noch "Experten" die glauben, auf Basis aktueller KI-Systeme (LLMs) sei eine Superintelligenz (AGI) erreichbar.
LLMs werden immer halluzinieren. Das ist systemimmanent.
LLMs sind leistungsfähige sprachliche Heuristikmaschinen. Mehr nicht.
Und es gibt immer noch "Experten" die glauben, auf Basis aktueller KI-Systeme (LLMs) sei eine Superintelligenz (AGI) erreichbar.
LLMs werden immer halluzinieren. Das ist systemimmanent.
LLMs sind leistungsfähige sprachliche Heuristikmaschinen. Mehr nicht.
glm 5 now has weights and gguf, we'll see about 2.5 :p
Not that anyone here can really run glm 5 locally anyways
Not that anyone here can really run glm 5 locally anyways
February 11, 2026 at 5:37 PM
glm 5 now has weights and gguf, we'll see about 2.5 :p
Not that anyone here can really run glm 5 locally anyways
Not that anyone here can really run glm 5 locally anyways
GLM-5 is live on AI Gateway | Vercel News
GLM-5 is live on AI Gateway - Vercel
You can now access Z.AI's latest model, GLM 5, via Vercel's AI Gateway with no other provider accounts required.
vercel.com
February 11, 2026 at 8:18 PM
GLM-5 is live on AI Gateway | Vercel News
z.ai increased price of coding plan by 33% but they also just gave an agent in a web environment so conflicted. Plus I've been using claude-opus-4.6 at work and GLM-5 is just as good. I mean it's really really good. Serving a 750B param model vs a 350B param model, costs might go up.
Z.ai - Free AI Chatbot & Agent powered by GLM-5 & GLM-4.7
Meet Z.ai, your free AI-powered assistant. Build websites, create slides, analyze data, and get instant answers. Fast, smart, and reliable, powered by GLM-5.
z.ai
February 11, 2026 at 10:02 PM
z.ai increased price of coding plan by 33% but they also just gave an agent in a web environment so conflicted. Plus I've been using claude-opus-4.6 at work and GLM-5 is just as good. I mean it's really really good. Serving a 750B param model vs a 350B param model, costs might go up.
Source: Chinese AI startup Zhipu anonymously released its new AI model on OpenRouter under the name Pony Alpha; Zhipu plans to launch GLM-5 later this week (Juro Osawa/The Information)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
February 10, 2026 at 10:35 AM
Source: Chinese AI startup Zhipu anonymously released its new AI model on OpenRouter under the name Pony Alpha; Zhipu plans to launch GLM-5 later this week (Juro Osawa/The Information)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
Core layer: as a workforce our programming labour is now disposable like the replicants are because any product person with a dream and a claude max subscription can now say "make the user's icon have party streamers when it's their birthday".
Hell, not even Claude Max, I mostly use GLM 4.7.
Hell, not even Claude Max, I mostly use GLM 4.7.
February 11, 2026 at 10:38 AM
Core layer: as a workforce our programming labour is now disposable like the replicants are because any product person with a dream and a claude max subscription can now say "make the user's icon have party streamers when it's their birthday".
Hell, not even Claude Max, I mostly use GLM 4.7.
Hell, not even Claude Max, I mostly use GLM 4.7.
Okay the fact that GLM-5 has the lowest hallucination rate really does standout as something genuinely interesting. That does lend to the idea it isn't benchmaxxed.
February 12, 2026 at 1:14 AM
Okay the fact that GLM-5 has the lowest hallucination rate really does standout as something genuinely interesting. That does lend to the idea it isn't benchmaxxed.
GLM just entered the OCR field🔥
huggingface.co/zai-org/GLM-...
✨ 0.9B
✨ MIT licensed
✨ Multimodal GLM-V architecture
✨ #1 on OmniDocBench v1.5 (94.62)
huggingface.co/zai-org/GLM-...
✨ 0.9B
✨ MIT licensed
✨ Multimodal GLM-V architecture
✨ #1 on OmniDocBench v1.5 (94.62)
February 3, 2026 at 8:10 AM
GLM just entered the OCR field🔥
huggingface.co/zai-org/GLM-...
✨ 0.9B
✨ MIT licensed
✨ Multimodal GLM-V architecture
✨ #1 on OmniDocBench v1.5 (94.62)
huggingface.co/zai-org/GLM-...
✨ 0.9B
✨ MIT licensed
✨ Multimodal GLM-V architecture
✨ #1 on OmniDocBench v1.5 (94.62)
I tried appling general linear models to the hits, kudos, and comments BTS fic receive on AO3, given factors of fic age, length, rating, completeness and primo ship. By modeling these factors together, we can see how much they affect work reception.
Details ao3.org/works/76528861/chapters/208145256
Details ao3.org/works/76528861/chapters/208145256
February 11, 2026 at 8:34 PM
I tried appling general linear models to the hits, kudos, and comments BTS fic receive on AO3, given factors of fic age, length, rating, completeness and primo ship. By modeling these factors together, we can see how much they affect work reception.
Details ao3.org/works/76528861/chapters/208145256
Details ao3.org/works/76528861/chapters/208145256
#statstab #479 Different ways of calculating OLS regression coefficients (in R)
Thoughts: The are many ways to skin a variable...
#rstats #regression #modelling #tutorial #r #glm #ols #coding
thomvolker.github.io/blog/2506_re...
Thoughts: The are many ways to skin a variable...
#rstats #regression #modelling #tutorial #r #glm #ols #coding
thomvolker.github.io/blog/2506_re...
Different ways of calculating OLS regression coefficients (in R) – Thom Volker
Many different ways of calculating OLS regression coefficients exist, but some ways are more efficient than others. In this post we discuss some of the most common ways of calculating OLS regression c...
thomvolker.github.io
February 5, 2026 at 8:16 PM
#statstab #479 Different ways of calculating OLS regression coefficients (in R)
Thoughts: The are many ways to skin a variable...
#rstats #regression #modelling #tutorial #r #glm #ols #coding
thomvolker.github.io/blog/2506_re...
Thoughts: The are many ways to skin a variable...
#rstats #regression #modelling #tutorial #r #glm #ols #coding
thomvolker.github.io/blog/2506_re...
z.ai is $10/month, and also has a free model in glm-4.7-flash; it's best thought of as a single-file code assistant rather than a project-level assistant, but I am using it right now for some stuff.
Z.ai - Free AI Chatbot & Agent powered by GLM-5 & GLM-4.7
Meet Z.ai, your free AI-powered assistant. Build websites, create slides, analyze data, and get instant answers. Fast, smart, and reliable, powered by GLM-5.
z.ai
February 12, 2026 at 12:39 AM
z.ai is $10/month, and also has a free model in glm-4.7-flash; it's best thought of as a single-file code assistant rather than a project-level assistant, but I am using it right now for some stuff.