6 open-weights AI models from smaller labs. The largest context window is 128K tokens (HyperCLOVA X). Newest check: 31 Oct 2026.
94 of 197 models were checked against the vendor's own page in the last 30 days.
| Model | Lab | Context | In / out per 1M | Reads and writes | Status | Checked |
|---|---|---|---|---|---|---|
| Intern-DecisionShanghai AI Lab's open-weights multimodal decision models (0.8B, 2B, 4B) return typed choices with probabilities in one pass, released Sept 2026. | Shanghai AI Lab | -- | -- | text, image → text | Preview | 18 Oct |
| LTX-2.5Lightricks' open-weights video and audio model that renders synchronized 4K HDR clips from text, image or video input, self-hostable and fine-tunable. | Lightricks | -- | -- | text, image, video → video, audio | Available | 15 Oct |
| Kanana-2 30B-A3B InstructOpen-weight mixture-of-experts chat model from Kakao, released January 2026, with 3B active parameters and six supported languages. | Kakao | 33K | -- | text → text | Available | 13 Oct |
| HyperCLOVA XNaver's HyperCLOVA X family runs from free SEED weights on Hugging Face to managed CLOVA Studio models, led by an open reasoning model that reads text, images and video. | Naver | 128K | -- | text, image, video → text | Available | 31 Oct |
| FLUX.2FLUX.2 by Black Forest Labs (Nov 2025) is a 32B rectified flow transformer family producing up to 4MP images. Five variants span open-weight klein to grounded [max], with multi-reference editing and hex color control. | Black Forest Labs | -- | -- | text, image → image | Available | 4 Oct |
| Osmosis-Structure-0.6BOpen-weight 0.6B model built on Qwen3 that extracts schema-valid structured output from reasoning traces, with 996 monthly Hugging Face downloads. | Osmosis | -- | -- | text → text | Available | 15 Sep |
A person checks every record against the vendor's own page. Rankings cannot be bought. How Hokai checks records. Not sure what you need? Describe the job to Smart Match and get three picks with reasons.