Hugging Face Pricing
Hugging Face charges between $0.250 and $4.40 per 1 million output tokens, depending on the model. Qwen3.5 9B is the cheapest at $0.250/1M output and $0.170/1M input; GLM-5.2 is the most expensive at $4.40/1M output. The largest context window is 1.0M tokens (DeepSeek V4 Flash). Prices are USD, current as of July 20, 2026.
Last updated · synced weekly from the upstream model catalog
Cheapest input
$0.070/1M
Qwen3-Coder 30B-A3B Instruct
Cheapest output
$0.250/1M
Qwen3.5 9B
Longest context
1.0M
DeepSeek V4 Flash
Models priced
48
Avg $1.85/1M output
Pricing by model
All prices in USD per 1 million tokens. All 48 priced models, cheapest first.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Qwen3.5 9B VisionReasoningTools | $0.170 | $0.250 | 262K |
| Qwen3-Coder 30B-A3B Instruct Tools | $0.070 | $0.260 | 262K |
| DeepSeek V4 Flash ReasoningTools | $0.140 | $0.280 | 1.0M |
| MiMo-V2-Flash ReasoningTools | $0.100 | $0.300 | 262K |
| Step 3.5 Flash ReasoningTools | $0.100 | $0.300 | 262K |
| DeepSeek-V3.2 ReasoningTools | $0.280 | $0.400 | 164K |
| Gemma 4 26B A4B IT VisionReasoningTools | $0.130 | $0.400 | 262K |
| Gemma 4 31B IT VisionReasoningTools | $0.140 | $0.400 | 262K |
| GPT OSS 20B ReasoningTools | $0.100 | $0.500 | 131K |
| Qwen3 32B ReasoningTools | $0.290 | $0.590 | 131K |
| GPT OSS 120B ReasoningTools | $0.250 | $0.690 | 131K |
| Llama-3.3-70B-Instruct Tools | $0.590 | $0.790 | 131K |
| Qwen3 235B-A22B ReasoningTools | $0.200 | $0.800 | 41K |
| GLM-4.5-Air ReasoningTools | $0.130 | $0.850 | 131K |
| DeepSeek V4 Pro ReasoningToolsCache | $0.435 | $0.870 | 1.0M |
| Qwen3.6 35B-A3B VisionReasoningTools | $0.150 | $0.950 | 262K |
| Qwen3-Next-80B-A3B-Instruct Tools | $0.250 | $1.00 | 262K |
| Step 3.7 Flash VisionReasoningTools | $0.200 | $1.15 | 262K |
| MiniMax-M2 ReasoningTools | $0.300 | $1.20 | 205K |
| MiniMax-M2.1 ReasoningTools | $0.300 | $1.20 | 205K |
| MiniMax-M2.5 ReasoningToolsCache | $0.300 | $1.20 | 205K |
| MiniMax-M2.7 ReasoningToolsCache | $0.300 | $1.20 | 205K |
| MiniMax-M3 VisionReasoningTools | $0.300 | $1.20 | 524K |
| Qwen3-Coder-Next Tools | $0.200 | $1.50 | 262K |
| GLM-4.5V VisionReasoningTools | $0.600 | $1.80 | 66K |
| Qwen3-Coder-480B-A35B-Instruct Tools | $2.00 | $2.00 | 262K |
| Qwen3-Next-80B-A3B-Thinking Tools | $0.300 | $2.00 | 262K |
| Qwen3.5 35B-A3B VisionReasoningTools | $0.250 | $2.00 | 262K |
| GLM-4.5 ReasoningTools | $0.600 | $2.20 | 131K |
| GLM-4.6 ReasoningTools | $0.550 | $2.20 | 205K |
| GLM-4.7 ReasoningToolsCache | $0.600 | $2.20 | 205K |
| Qwen3.5 27B VisionReasoningTools | $0.300 | $2.40 | 262K |
| DeepSeek-R1 ReasoningTools | $0.700 | $2.50 | 64K |
| Kimi-K2-Thinking ReasoningToolsCache | $0.600 | $2.50 | 262K |
| Kimi-K2-Instruct Tools | $1.00 | $3.00 | 131K |
| Kimi-K2-Instruct-0905 Tools | $1.00 | $3.00 | 262K |
| Kimi-K2.5 VisionReasoningToolsCache | $0.600 | $3.00 | 262K |
| MiMo-V2.5-Pro ReasoningTools | $1.00 | $3.00 | 1.0M |
| Qwen3-235B-A22B-Thinking-2507 ReasoningTools | $0.300 | $3.00 | 262K |
| Qwen3.6 27B VisionReasoningTools | $0.470 | $3.19 | 262K |
| GLM-5 ReasoningToolsCache | $1.00 | $3.20 | 203K |
| GLM-5.1 ReasoningToolsCache | $1.00 | $3.20 | 203K |
| Qwen3.5 122B-A10B VisionReasoningTools | $0.400 | $3.20 | 262K |
| Qwen3.5-397B-A17B VisionReasoningTools | $0.600 | $3.60 | 262K |
| Kimi K2.7 Code VisionReasoningTools | $0.950 | $4.00 | 262K |
| Kimi-K2.6 VisionReasoningToolsCache | $0.950 | $4.00 | 262K |
| GLM-5.2 ReasoningTools | $1.40 | $4.40 | 262K |
| DeepSeek-R1-0528 ReasoningTools | $3.00 | $5.00 | 164K |
Frequently asked questions
How much does Hugging Face cost per 1M tokens?
Hugging Face pricing ranges from $0.250 to $4.40 per 1 million output tokens across 48 models. Input tokens are cheaper than output on every model. Rates current as of July 20, 2026.
What is the cheapest Hugging Face model?
Qwen3.5 9B is the cheapest Hugging Face model on output tokens at $0.250 per 1M, while Qwen3-Coder 30B-A3B Instruct is cheapest on input at $0.070 per 1M. Which wins for you depends on your input-to-output ratio.
What is the largest Hugging Face context window?
DeepSeek V4 Flash has the largest context window in the Hugging Face catalog at 1.0M input tokens, with up to 384K output tokens per response.
Does Hugging Face support prompt caching?
Yes. DeepSeek V4 Pro reads cached input at $0.0036 per 1M tokens, against $0.435 per 1M uncached. Caching pays off when a long system prompt or document is reused across many requests.
Which Hugging Face models support reasoning or vision?
The Hugging Face catalog includes 40 reasoning models, 15 vision models, 48 with tool calling. Reasoning models bill their internal thinking as output tokens, so they cost more per visible response than the headline rate suggests.
How do I calculate my actual Hugging Face bill?
Multiply your input tokens by the input rate and your output tokens by the output rate, both per million, then add retries and any conversation history resent on each turn. Calcaas does this against your real usage pattern and shows the margin left at your price point.
Other providers
What will this actually cost you?
Per-token rates only tell you half the story. Calcaas multiplies them by your real usage — prompt size, response length, retries and conversation history — then shows the margin left at your price point.
Open the free cost calculator