gpt-oss-20b
Work through multi-step text problems on your own machine.
Find downloadable models for offline transcription, image generation, document search and coding. Compare use cases before choosing weights for your hardware.
Choose Whisper to turn speech into text or Kokoro to read text aloud. For searching your own files, BGE-M3 can turn passages into embeddings. For image generation, start with FLUX.1-schnell.
Try it. For transcription, use one minute of audio with a known transcript. Compare names, numbers and omissions, then measure how long your machine takes.
Before you commit. Open the model card first. Check its licence, download size, runtime and memory requirements for the precision you plan to use. A free download can still require substantial hardware.
15 entries
Work through multi-step text problems on your own machine.
Try multilingual conversations with a compact instruction model.
Generate a function, explain unfamiliar code or suggest a repair.
Ask questions about a photo using a smaller multimodal model.
Experiment with a distilled reasoning model on maths and logic prompts.
Generate image variations from text in a small number of sampling steps.
Transcribe a recorded interview into text for editing.
Retrieve related passages across languages for a document search tool.
Caption a scene or identify objects and their positions.
Turn written text into synthesized speech.
Generate illustrations locally with a widely supported diffusion model.
Isolate an object in an image or follow its mask through video.
Combine text and image inputs for document questions.
Find similar sentences using compact English-language embeddings.
Read visual details and answer questions about images.