Tuesday, October 13
4 new models on Hugging Face that day.
-
Edits imagesblack-forest-labs
flux-3-image-klein
A 5.6B image model from Black Forest Labs that makes images from text and edits them, with up to 10 reference images.
- One model for text-to-image, single-image edits and multi-reference compositions
- Returns one image per request; the Qwen3-VL-4B-Instruct text encoder comes separately
- Distilled checkpoint; for fine-tuning or LoRA training use the separate Base model
Minimum 12GB graphics card e.g. GeForce RTX 5060 Ti 16GB
Check the licence before useDownload needs the maker’s approval
Download -
Chat and writinginclusionAI
Ling-3.1-flash
A 560B hybrid reasoning model from inclusionAI, aimed at real-world work like office tasks, search and coding.
- The card reports gains on agent tasks such as office work, search and software engineering
- 560B parameters in total, only 25B active per token, with a 1M-token context
- Very large in total size, and the card lists no specific supported languages
Minimum Server with several graphics cards
Free for commercial use
Download -
Chat and writingsabicap
sabi-noema
A language model of about 27B from Sabi, built from Qwen3.8-27B with extra training.
- The card says it makes up answers less often than the base model: 15.1% vs 30.4%
- Dense model with a 262K-token context, weights in BF16
- The numbers come from the maker's own tests
Minimum 24GB graphics card e.g. GeForce RTX 5090
Free for commercial use
Download -
Reads imagesrespanai
span-02
A decision model that does not write text: it reads a situation and a question, then gives a probability for each answer.
- Takes text or JSON plus up to 4 images; answers are a choice, yes/no, a score or a number
- A fine-tune of Gemma 4 26B-A4B, about 4B active parameters, with its own engine included
- It cannot be loaded with transformers or vLLM; only the bundled engine runs it
Minimum 8GB graphics card, or 16GB computer memory (slower) e.g. GeForce RTX 5050
Free for commercial use
Download
The minimum is our estimate from the model’s size: a 4-bit copy for text models, 8-bit for image, video and sound. Licences and sizes come from each maker’s page; read the licence before using a model in your work.
