DeepSeek-V3.1
By deepseek-ai. 684.5 billion parameters, a context window of 163,840 tokens and the licence mit.
Facts
- Released
- 2025-08-21 the day the repository was first published on Hugging Face Hugging Face, read
- Licence
- mit the licence the model card declares Hugging Face, read
- Open weights
- Yes the weights are published in this Hugging Face repository Hugging Face, read
- Parameters
- 684.5 billion counted from the safetensors weight files Hugging Face, read
- Active parameters
- 51.1 billion estimate estimated from the expert counts and sizes in the configuration model configuration, read
- Knowledge cutoff
- Unknown
- Context window
- 163,840 tokens the maximum position embeddings in the model configuration model configuration, read
- Longest output
- Unknown
- Tool calling
- No the chat template in the tokenizer configuration has no place for tool definitions Hugging Face, read
- Structured output
- Unknown
- Reasoning controls
- Unknown
- Inputs and outputs
- text in, text out Hugging Face, read
- Good for
- No source names a use.
Prices
| Provider | Input | Output | Context | As of | Source |
|---|---|---|---|---|---|
| NanoGPT | 0.200 USD | 0.700 USD | 128,000 | older than 30 days; check the provider | models.dev, read |
| submodel | 0.200 USD | 0.800 USD | 75,000 | older than 30 days; check the provider | models.dev, read |
| Deep Infra | 0.250 USD | 0.950 USD | 163,840 | older than 30 days; check the provider | models.dev, read |
| Hugging Face | 0.270 USD | 1.00 USD | 131,072 | older than 30 days; check the provider | models.dev, read |
| Meganova | 0.270 USD | 1.00 USD | 164,000 | older than 30 days; check the provider | models.dev, read |
| SiliconFlow | 0.270 USD | 1.00 USD | 164,000 | older than 30 days; check the provider | models.dev, read |
| Baseten | 0.500 USD | 1.50 USD | 164,000 | older than 30 days; check the provider | models.dev, read |
| CoreWeave | 0.550 USD | 1.65 USD | 161,000 | older than 30 days; check the provider | models.dev, read |
| Pioneer | 0.560 USD | 1.68 USD | 163,840 | older than 30 days; check the provider | models.dev, read |
| Together AI | 0.600 USD | 1.70 USD | 131,072 | LiteLLM, read | |
| replicate | 0.672 USD | 2.02 USD | 163,840 | LiteLLM, read |
Run it on your own hardware
Weights are the sizes of the files a source lists, or an estimate from the parameter count where none does. Memory is an estimate: weights plus KV cache plus 512 MiB and 5 percent of the weights for runtime buffers. Check it against your hardware.
| Quantization | Weights | Memory at 8,192 tokens | Memory at 32,768 tokens |
|---|---|---|---|
| UD-TQ1_0 | 159 GiB | 168 GiB | 169 GiB |
| UD-IQ1_S | 179 GiB | 189 GiB | 191 GiB |
| UD-IQ1_M | 193 GiB | 203 GiB | 205 GiB |
| UD-IQ2_XXS | 210 GiB | 222 GiB | 224 GiB |
| UD-IQ2_M | 219 GiB | 231 GiB | 233 GiB |
| Q2_K | 229 GiB | 241 GiB | 243 GiB |
| Q2_K_L | 229 GiB | 242 GiB | 243 GiB |
| UD-Q2_K_XL | 238 GiB | 251 GiB | 253 GiB |
| UD-IQ3_XXS | 260 GiB | 275 GiB | 276 GiB |
| Q3_K_S | 271 GiB | 285 GiB | 287 GiB |
| UD-Q3_K_XL | 279 GiB | 294 GiB | 296 GiB |
| Q3_K_M | 298 GiB | 314 GiB | 316 GiB |
| IQ4_XS | 334 GiB | 351 GiB | 353 GiB |
| IQ4_NL | 353 GiB | 372 GiB | 373 GiB |
| Q4_0 | 354 GiB | 373 GiB | 374 GiB |
| Q4_K_S | 355 GiB | 374 GiB | 375 GiB |
| UD-Q4_K_XL | 360 GiB | 379 GiB | 381 GiB |
| Q4_K_M | 378 GiB | 397 GiB | 399 GiB |
| Q4_1 | 392 GiB | 413 GiB | 414 GiB |
| Q5_K_S | 431 GiB | 453 GiB | 455 GiB |
| Q5_K_M | 443 GiB | 467 GiB | 468 GiB |
| UD-Q5_K_XL | 451 GiB | 475 GiB | 477 GiB |
| Q6_K | 513 GiB | 540 GiB | 542 GiB |
| UD-Q6_K_XL | 535 GiB | 563 GiB | 564 GiB |
| Q8_0 | 664 GiB | 699 GiB | 700 GiB |
| UD-Q8_K_XL | 727 GiB | 764 GiB | 766 GiB |
| BF16 | 1,250 GiB | 1,314 GiB | 1,315 GiB |
Start it with a local runtime
- Ollama
ollama run hf.co/unsloth/DeepSeek-V3.1-GGUF:Q4_0 - LM Studio
lms get unsloth/DeepSeek-V3.1-GGUF lms server start - llama.cpp server
llama-server -hf unsloth/DeepSeek-V3.1-GGUF:Q4_0 --jinja - vLLM
vllm serve deepseek-ai/DeepSeek-V3.1 - SGLang
sglang serve --model-path deepseek-ai/DeepSeek-V3.1 --port 30000
Use it from your harness
Set up for Baseten with the model deepseek-ai/DeepSeek-V3.1. Each endpoint page has the same setup for its own address.
OpenCode
Put this in opencode.json in your project folder:
{
"$schema": "https://opencode.ai/config.json",
"provider": {
"baseten": {
"npm": "@ai-sdk/openai-compatible",
"name": "Baseten",
"options": {
"baseURL": "https://inference.baseten.co/v1",
"apiKey": "{env:BASETEN_API_KEY}"
},
"models": {
"deepseek-ai/DeepSeek-V3.1": {
"name": "DeepSeek-V3.1"
}
}
}
}
}- OpenCode reads any OpenAI-compatible address through the @ai-sdk/openai-compatible package, and an address that speaks the Responses API through @ai-sdk/openai.
From OpenCode documentation, read .
Pi
Put this in ~/.pi/agent/models.json:
{
"providers": {
"baseten": {
"baseUrl": "https://inference.baseten.co/v1",
"api": "openai-completions",
"apiKey": "$BASETEN_API_KEY",
"models": [
{
"id": "deepseek-ai/DeepSeek-V3.1"
}
]
}
}
}- The apiKey field can name an environment variable as $NAME.
From Pi documentation, read .
Codex
Codex speaks only the Responses API, and Baseten documents no Responses address. A gateway that offers one can sit in between.
- Codex speaks the Responses API only: responses is the one supported wire API of a custom provider. Ollama and LM Studio are built in and start with --oss.
From Codex documentation, read .
Claude Code
Claude Code sends Anthropic Messages requests, and Baseten documents no such address. A gateway that translates to that API can sit in between.
- Claude Code sends Anthropic Messages requests to ANTHROPIC_BASE_URL. Anthropic says it does not support routing Claude Code to models other than Claude through any gateway, so some features may not work with another model.
From Claude Code documentation, read .
Other ways to reach it
- DeepInfra, model
deepseek-ai/DeepSeek-V3.1 - Hugging Face, model
deepseek-ai/DeepSeek-V3.1 - Meganova, model
deepseek-ai/DeepSeek-V3.1 - NanoGPT, model
deepseek-ai/DeepSeek-V3.1 - Pioneer, model
deepseek-ai/DeepSeek-V3.1 - SiliconFlow, model
deepseek-ai/DeepSeek-V3.1 - submodel, model
deepseek-ai/DeepSeek-V3.1 - Together AI, model
deepseek-ai/DeepSeek-V3.1 - CoreWeave, model
deepseek-ai/DeepSeek-V3.1 - Ollama, model
hf.co/unsloth/DeepSeek-V3.1-GGUF:Q4_0
Published results
These are results other people published. Baltor did not run them and does not rank models by them.
- Results the maker published on the model card published by deepseek-ai Hugging Face, read
Two more publishers are linked, never copied, because their terms do not allow it:
- Artificial Analysis publishes its own intelligence, speed and price measurements.
- OpenRouter lists current prices and uptime for each provider it routes to.
Sources of this page
- huggingface.co/api/models/deepseek-ai/DeepSeek-V3.1, read
- huggingface.co/deepseek-ai/DeepSeek-V3.1/raw/main/config.json, read
- huggingface.co/api/models/unsloth/DeepSeek-V3.1-GGUF/tree/main, read
- models.dev/api.json, read
- github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json, read
Paid links
No link in this directory is a paid link or an ad, and no listing is paid for. The order and the contents of every list come from the sources named on this page.