Models

Discover the base models available for fine-tuning and serving, with per-model limits and pricing.

Models are the open-weight bases you fine-tune and serve. The catalogue spans language, vision, embedding and reranking, image generation, video generation and audio models, and is where an integration discovers model IDs and each model's limits: context length, the highest LoRA rank it accepts, whether it can be trained at all, and when it will be withdrawn. Pass a model ID from here to fine-tuning jobs, training runs and the tokenization routes.

Every model is metered in tokens. For image, video and audio models, those are the tokens of the encoded media: a megapixel image is roughly 4K tokens, a five-second 720p clip roughly 70K, and a minute of audio roughly 3K. Embedding and reranking models are metered on the tokens they read, because they return a vector or a score rather than tokens.

Model IDs are namespaced strings such as Qwen/Qwen3-8B and meta-llama/Llama-3.3-70B.

List base models

GET/v1/modelsscope: read

Returns the catalogue of base models. Use the filters to narrow the list to models you can actually use: trainable for fine-tuning targets, modality for the modalities your application works in.

Query parameters

trainableboolean

Filter to models that can be fine-tuned.

modalitystring

Filter by modality: text, image, audio, video, embedding or rerank. Matches models that accept it as input or produce it as output. embedding returns the bi-encoders and rerank the cross-encoders, which score a query against a document instead of returning a vector.

Each entry carries context_length (the maximum tokens per sequence), input_modalities and output_modalities (the types the model accepts and produces), max_lora_rank (the highest LoRA rank accepted for that model, null when trainable is false and the model can only be served) and deprecates_at. When deprecates_at is set, the model will be withdrawn at that time; integrations should surface a warning well before the date rather than fail on it. null means no withdrawal is scheduled.

curl "https://api.nucleus-ai.io/v1/models?trainable=true" \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

200 OK

{
  "data": [
    {
      "id": "nvidia/Nemotron-3.5-Lightning",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "openai/gpt-oss-20b",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "nvidia/Nemotron-3-Nano-30B-A3B",
      "context_length": 65536,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-8B",
      "context_length": 32768,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3.8-27B-A3B",
      "context_length": 262144,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3.6-35B-A3B",
      "context_length": 262144,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "zai-org/GLM-5.3-Flash",
      "context_length": 1048576,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "openai/gpt-oss-120b",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "MiniMaxAI/MiniMax-M3",
      "context_length": 1048576,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "mistralai/Mistral-Medium-3.5",
      "context_length": 262144,
      "input_modalities": ["text", "image"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "deepseek-ai/DeepSeek-V4-Flash-0731",
      "context_length": 1048576,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "deepseek-ai/DeepSeek-V4-Flash",
      "context_length": 1048576,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": false,
      "max_lora_rank": null,
      "deprecates_at": "2027-01-31T00:00:00Z"
    },
    {
      "id": "meta-llama/Llama-3.3-70B",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "deepseek-ai/DeepSeek-V3.2",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "deepseek-ai/DeepSeek-V3.1",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": "2027-01-31T00:00:00Z"
    },
    {
      "id": "zai-org/GLM-5.2",
      "context_length": 1048576,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "moonshotai/Kimi-K2.6",
      "context_length": 131072,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": "2027-01-31T00:00:00Z"
    },
    {
      "id": "zai-org/GLM-5.3",
      "context_length": 1048576,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "nvidia/Nemotron-3-Ultra-550B-A55B",
      "context_length": 65536,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "deepseek-ai/DeepSeek-V4-Pro",
      "context_length": 1048576,
      "input_modalities": ["text"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 16,
      "deprecates_at": null
    },
    {
      "id": "moonshotai/Kimi-K3",
      "context_length": 1048576,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": false,
      "max_lora_rank": null,
      "deprecates_at": null
    },
    {
      "id": "google/gemma-4-E4B",
      "context_length": 131072,
      "input_modalities": ["text", "image", "video", "audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-VL-30B-A3B-Instruct",
      "context_length": 131072,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-VL-8B-Instruct",
      "context_length": 131072,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3.5-9B",
      "context_length": 65536,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "google/gemma-4-31B",
      "context_length": 262144,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
      "context_length": 131072,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3.5-397B-A17B",
      "context_length": 65536,
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "google/embeddinggemma-300m",
      "context_length": 2048,
      "input_modalities": ["text"],
      "output_modalities": ["embedding"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "BAAI/bge-reranker-v2-m3",
      "context_length": 8192,
      "input_modalities": ["text"],
      "output_modalities": ["rerank"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "BAAI/bge-en-icl",
      "context_length": 32768,
      "input_modalities": ["text"],
      "output_modalities": ["embedding"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-Embedding-8B",
      "context_length": 32768,
      "input_modalities": ["text"],
      "output_modalities": ["embedding"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "nomic-ai/nomic-embed-code",
      "context_length": 32768,
      "input_modalities": ["text"],
      "output_modalities": ["embedding"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-Reranker-8B",
      "context_length": 32768,
      "input_modalities": ["text"],
      "output_modalities": ["rerank"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "black-forest-labs/FLUX.2-klein",
      "context_length": 8192,
      "input_modalities": ["text", "image"],
      "output_modalities": ["image"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "stabilityai/stable-diffusion-3.5-large",
      "context_length": 8192,
      "input_modalities": ["text", "image"],
      "output_modalities": ["image"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen-Image",
      "context_length": 16384,
      "input_modalities": ["text"],
      "output_modalities": ["image"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "black-forest-labs/FLUX.2-dev",
      "context_length": 16384,
      "input_modalities": ["text", "image"],
      "output_modalities": ["image"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "tencent/HunyuanVideo-1.5",
      "context_length": 131072,
      "input_modalities": ["text", "image"],
      "output_modalities": ["video"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Lightricks/LTX-2",
      "context_length": 262144,
      "input_modalities": ["text", "image"],
      "output_modalities": ["video", "audio"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "Wan-AI/Wan2.2-T2V-A14B",
      "context_length": 131072,
      "input_modalities": ["text"],
      "output_modalities": ["video"],
      "trainable": true,
      "max_lora_rank": 32,
      "deprecates_at": null
    },
    {
      "id": "openai/whisper-large-v3-turbo",
      "context_length": 1500,
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "nvidia/parakeet-tdt-1.1b",
      "context_length": 8192,
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "nvidia/diar_sortformer_4spk-v1",
      "context_length": 8192,
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "nvidia/Nemotron-3.5-ASR-Multilingual",
      "context_length": 8192,
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "Qwen/Qwen3-TTS-1.7B",
      "context_length": 4096,
      "input_modalities": ["text"],
      "output_modalities": ["audio"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "openai/whisper-large-v3",
      "context_length": 1500,
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "canopylabs/orpheus-3b-0.1-ft",
      "context_length": 4096,
      "input_modalities": ["text"],
      "output_modalities": ["audio"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    },
    {
      "id": "mistralai/Voxtral-4B-TTS-2603",
      "context_length": 4096,
      "input_modalities": ["text"],
      "output_modalities": ["audio"],
      "trainable": true,
      "max_lora_rank": 64,
      "deprecates_at": null
    }
  ]
}

Retrieve a base model

GET/v1/models/{model_id}scope: read

Returns one model with its full limits and pricing. chat_template tells you whether Render a conversation can apply a chat template for this model; when it is false, tokenize your data yourself with Tokenize text.

Path parameters

model_idstringrequired

ID of the model, including its namespace, e.g. Qwen/Qwen3-8B.

price_per_mtok_train and price_per_mtok_infer are prices per million tokens for training and inference. Either can be null when a per-token price is not published for the model.

curl https://api.nucleus-ai.io/v1/models/Qwen/Qwen3-8B \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

200 OK

{
  "id": "Qwen/Qwen3-8B",
  "context_length": 32768,
  "input_modalities": ["text"],
  "output_modalities": ["text"],
  "tokenizer": "qwen3",
  "chat_template": true,
  "trainable": true,
  "max_lora_rank": 64,
  "price_per_mtok_train": 0.4,
  "price_per_mtok_infer": 0.3,
  "deprecates_at": null
}

Models

List base models

Retrieve a base model

On this page