Developers Are Rediscovering e5-small-v2's 33M-Parameter Search Power
A 33M parameter English embedding model is climbing Hugging Face trending charts, offering 384 dimensional vectors with surprisingly competitive MTEB scores.
20 stories
A 33M parameter English embedding model is climbing Hugging Face trending charts, offering 384 dimensional vectors with surprisingly competitive MTEB scores.
Snowflake's open-source Arctic Embed L delivers 55.98 NDCG@10 on MTEB Retrieval, matching closed APIs from OpenAI and Cohere at a fraction of the size.
Project Maya runs the 321B GLM-5.3-Flash mixture-of-experts model on one or two consumer NVIDIA GPUs by tiering experts across VRAM, RAM, and SSD.
A patched serving stack runs the 27B Ternary Bonsai 2 at the full 262k context window on a single 12 GB consumer GPU, with agentic tool-calling finally fixed.
A new camera-control adapter for Qwen-Image-2.1 lets you reshoot any photo from 72 preset framings using a simple prompt grammar.
LAION's open ViT-B/16 CLIP, trained on 2 billion image-text pairs, hits 70.2% zero-shot on ImageNet and powers nearly 100 spaces.
Unsloth released GGUF quantizations of Tongyi's Z-Image-Turbo, a 6B parameter text-to-image model that runs in under 16GB VRAM with 8-step inference.
A locally-run 12B model fine-tuned from Gemma-4 rewrites AI drafts into natural human prose while preserving every number, name, and quote.
A vision-capable 27B decision model that returns calibrated probabilities in one forward pass, matching collaborative filtering with zero training data.
A compact SigLIP2 fine-tune hits 97.2% accuracy on binary gender classification, pulling 778K downloads with a 93M-parameter footprint.
A new vision backbone rewrites its own learning rules while scanning an image, hitting strong ImageNet, COCO, and ADE20K numbers with linear-time cost.
ByteShape ships ShapeLearn-quantized GGUFs of Qwen3.8-27B that fit on 12 GB GPUs and hit 176 tokens per second on an RTX 5090.
A ternary-quantized 27B vision-language model shrinks to 7.3 GB and specializes in writing ComfyUI image and video generation prompts.
A community post-trained Qwen variant uses RLOO to stop endless re-derivation, cutting 94K-token truncations and pushing LiveCodeBench from 83 to 90.
A Chinese Telecom lab released a 29B mixture-of-experts model with only 4B active parameters, 256K context, and strong agentic coding benchmarks.
Aleph Alpha drops a 78B mixture-of-experts reasoning model with 3.46B active parameters, 1M-token context, and Apache 2.0 weights, tuned for German-English sovereign deployments.
Red Hat AI shipped an FP8 build of NVIDIA's hybrid Mamba-MoE Lightning model, halving memory while keeping the 1M-token agent workhorse on a single GPU.
OpenBMB released a 324M-parameter draft model that accelerates MiniCPM5-2B inference using semi-autoregressive speculative decoding, hitting 5.5x accepted length on average.
A compact 44M-parameter encoder from Microsoft Research keeps climbing the trending charts, offering ELECTRA-style efficiency and strong NLU accuracy at tiny inference cost.
A from-scratch Vietnamese TTS model ships 48 kHz speech, 23 preset voices, instant cloning, and 16 concurrent real-time streams on a single RTX 3060.