Skip to main content
Create a new model from a Modelfile, GGUF file, or safetensors format. This endpoint supports streaming progress updates during model creation.

​
Request

​
Endpoint

POST /api/create

​
Request Body

​
model
string
required
Name for the new model (e.g., my-model, namespace/model:tag)
​
from
string
Model name to use as base, or path to local file
​
files
object
Map of file paths to blob digests for model files (GGUF or safetensors)Example: {"model.gguf": "sha256:abc123...", "config.json": "sha256:def456..."}
​
adapters
object
Map of LoRA adapter files to include when creating the model
​
stream
boolean
default:"true"
Enable streaming of progress updates
​
quantize
string
Quantization format for the model (e.g., "Q4_0", "Q5_K_M", "Q8_0")
​
remote_host
string
URL of upstream Ollama API for remote models
​
template
string
Custom prompt template to use for the model
​
system
string
System prompt for the model
​
license
string | array
License information (string or array of strings)
​
parameters
object
Map of model parameters (e.g., temperature, top_k, etc.)
​
messages
array
List of messages to embed in the model
​
renderer
string
Template renderer to use
​
parser
string
Output parser to use (e.g., "harmony" for GPT-OSS models)
​
requires
string
Minimum Ollama version required by the model
​
info
object
Additional metadata for the model (not exposed in Modelfiles)

​
Response

​
Response Fields

​
status
string
Status message describing the current operation
​
digest
string
Digest of the layer being processed (during operations)
​
total
integer
Total size in bytes (during file operations)
​
completed
integer
Bytes completed (during file operations)

​
Examples

​
Create from Base Model

curl http://localhost:11434/api/create -d '{
  "model": "my-custom-model",
  "from": "llama3.2",
  "system": "You are a helpful coding assistant.",
  "parameters": {
    "temperature": 0.7,
    "top_p": 0.9
  }
}'

​
Create from GGUF File

First, upload the GGUF file: 
# Upload the file and get its digest
digest=$(sha256sum model.gguf | awk '{print "sha256:" $1}')
curl http://localhost:11434/api/blobs/$digest --data-binary @model.gguf
Then create the model: 
curl http://localhost:11434/api/create -d '{
  "model": "my-gguf-model",
  "files": {
    "model.gguf": "'$digest'"
  }
}'

​
Create with Quantization

curl http://localhost:11434/api/create -d '{
  "model": "my-quantized-model",
  "from": "llama3.2",
  "quantize": "Q4_K_M"
}'

​
Create from Safetensors

curl http://localhost:11434/api/create -d '{
  "model": "my-st-model",
  "files": {
    "model-00001-of-00002.safetensors": "sha256:abc123...",
    "model-00002-of-00002.safetensors": "sha256:def456...",
    "config.json": "sha256:ghi789...",
    "tokenizer.json": "sha256:jkl012..."
  }
}'

​
Non-Streaming Response

curl http://localhost:11434/api/create -d '{
  "model": "my-model",
  "from": "llama3.2",
  "stream": false
}'

​
Example Streaming Response

{"status":"parsing modelfile"}
{"status":"looking up model manifest"}
{"status":"creating model layer"}
{"status":"writing manifest"}
{"status":"success"}

​
Error Responses

​
error
string
Description of the error

​
Common Errors

  • 400 Bad Request: Invalid model name, missing required fields, or invalid file paths
  • 404 Not Found: Base model not found
  • 500 Internal Server Error: Error during model creation
You must provide either from (base model) or files (model files), but not both.
When creating from files, ensure all files are uploaded via /api/blobs/:digest before calling /api/create.