Z.ai debuts open source GLM-4.6V, a native tool-calling vision model for multimodal reasoning
Fyra's Brief
Z.ai has released an open-source vision-language model series, GLM-4.6V, for multimodal reasoning and frontend automation. The model achieves state-of-the-art results across various benchmarks, including general VQA, chart understanding, and STEM reasoning.
Why it matters
The GLM-4.6V series represents a significant advancement in open-source multimodal AI and its potential applications in various industries, including healthcare, finance, and education.
Community talk
Cruxy: Train 1.5B models on 4GB VRAM - new optimiser just released
DeepSeek-V3.2-REAP: 508B and 345B checkpoints
Anthropic is donating the Model Context Protocol (MCP) to the Linux Foundation
Anthropic hands over "Model Context Protocol" (MCP) to the Linux Foundation — aims to establish Universal Open Standard for Agentic AI
GLM-4.6V AWQ is released
GLM-4.6V, the latest open-source vision language models
mbzuai ifm releases Open 70b model - beats qwen-2.5
Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length
The Best Open-Source 8B-Parameter LLM Built in the USA
Mistral 3 Large 675B up on huggingface
is the new Deepseek v3.2 that bad?
Frozen networks show usable early-layer intent: 1370× fewer FLOPs and 10× faster inference (code + weights)9
Tiny-A2D: An Open Recipe to Turn Any AR LM into a Diffusion LM
zai-org/GLM-4.6V-Flash (9B) is here
GLM-4.6 Derestricted
Unimpressed with Mistral Large 3 675B
Deepseek R1 671b Q4_K_M
Aquif 3.5 Max 1205 (42B-A3B)
VibeVoice Realtime 0.5B - OpenAI Compatible /v1/audio/speech TTS Server
"Router mode is experimental" | llama.cpp now has a router mode and I didn't know.
Qwen3-TTS
[P] 96.1M Rows of iNaturalist Research-Grade plant images (with species names)
Open Unified TTS - Turn any TTS into an unlimited-length audio generator
Mistral 3 14b against the competition ?
Qwen3-Next-80B-A3B or Gpt-oss-120b?
[P] Zero Catastrophic Forgetting in MoE Continual Learning: 100% Retention Across 12 Multimodal Tasks (Results + Reproducibility Repo)
VLLM v0.12.0 supports NVFP4 for SM120 (RTX 50xx and RTX PRO 6000 Blackwell)
The "Confident Idiot" Problem: Why LLM-as-a-Judge fails in production.
apple/CLaRa-7B-Instruct · Hugging Face
I built an open-source prompt layering system after LLMs kept ignoring my numerical weights
Robert now supports full ChatGPT embodiment. You can switch seamlessly between manual and AI control, and I’ll be releasing the entire system as open source. More info in my profile, and in comments.
anyone else actually impressed with haiku 4.5?
Multi-agent orchestration is the future of AI coding. Here are some OSS tools to check out.
Open-sourced a collection of Claude skills
I ran Claude Code in a self-learning loop until it succesfully translated our entire Python repo to TypeScript
I built a "Guardrails-First" Agent template for the OpenAI SDK (Open Source).
[open source] I finetuned my own LLM in 20m on my personal notes. Now it thinks in my style.
[Resource] 20,000+ Pages of U.S. House Oversight Epstein Estate Docs (OCR'd & Cleaned for RAG/Analysis)
[P] I trained Qwen2.5-Coder-7B for a niche diagramming language and reached 86% code accuracy
I trained a 7B to learn a niche language and reaching 86% code accuracy
I built a Mistral inference engine from scratch