← back to Exo

resources/inference_model_cards/mlx-community--DeepSeek-V4-Flash.toml

22 lines

model_id = "mlx-community/DeepSeek-V4-Flash"
n_layers = 43
hidden_size = 4096
num_key_value_heads = 1
supports_tensor = true
tasks = ["TextGeneration"]
family = "deepseek"
quantization = "8bit"
base_model = "DeepSeek V4 Flash"
capabilities = ["text", "thinking", "thinking_toggle"]
reasoning_dialect = "tool_conditional"

context_length = 1048576
backends = ["MlxMetal", "MlxCuda", "MlxCpu"]
[storage_size]
in_bytes = 155095760030

# Source: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
[sampling_defaults]
temperature = 1.0
top_p = 1.0