modelai4bharatIndicConformer multilingualSpeech-recognition checkpoint from AI4Bharat. Real deployment needs dialect, audio-quality and domain-specific error analysis.Speech recognitionIndian sourceIndian languagesHugging FaceMIT
modelai4bharatIndicTrans2 English → IndicDistilled translation checkpoint for English-to-Indic workflows. Review publisher access conditions and evaluate each target language separately.TranslationIndian sourceIndian languagesHugging FaceMIT
modelHuggingFaceTBSmolLM2 1.7B InstructSmall instruction-following language model. Useful for building a reproducible local baseline with clear checkpoint provenance.Text generationHugging FaceApache 2.0
modelQwenQwen3 Embedding 0.6BEmbedding checkpoint for retrieval experiments. Evaluate retrieval relevance on the documents and languages used by the application.Embeddings / feature extractionHugging FaceApache 2.0
modelQwenQwen2.5 Coder 1.5B InstructInstruction-tuned coding checkpoint for investigating code-focused tasks. Published capabilities are not independent Suvyo measurements.Text generationHugging FaceApache 2.0
modelQwenQwen3.5 0.8BCompact Qwen3.5 checkpoint. Inspect the model card, modalities and runtime compatibility before choosing an inference stack.Image and text understandingHugging FaceApache 2.0
modelQwenQwen3 1.7BText-generation checkpoint in the Qwen3 family. A candidate for controlled local inference and structured-output experiments.Text generationHugging FaceApache 2.0
modelQwenQwen3 0.6BCompact text-generation checkpoint for testing small-model workflows. Compare actual task accuracy and runtime costs before deployment.Text generationHugging FaceApache 2.0
modelBAAIBGE-M3Retrieval model from BAAI for document-search experiments. Pair model selection with a held-out retrieval set and source attribution.Sentence similarityHugging FaceMIT