Hugging Face

Hugging Face offers free or low-cost inference for many premier open-source models via its Inference Providers ecosystem. Because many of these models have a generous free tier, this provider is fantastic for early prototyping, experimentation, and budget-friendly coding.

Website: https://huggingface.co/

Getting an API Key

1. Sign Up or Log In: Navigate to Hugging Face and log in to your account.
2. Go to Settings: Open your profile options and find the Access Tokens page.
3. Create a New Token: Issue a new access token, ensuring it has the correct permissions for inference.
4. Save the Token: Copy the token text right away and keep it secure.

Supported Models

Sypha works with these top Hugging Face models (which are currently free to use):

• moonshotai/Kimi-K2-Instruct (Default) - Top-tier reasoning model featuring a 131K context window, great for logic and coding tasks
• openai/gpt-oss-120b - Powerful 120B open-weight reasoning model capable of handling extreme complexity (131K context)
• openai/gpt-oss-20b - Fast 20B open-weight model that balances capabilities and accessibility perfectly (131K context)
• deepseek-ai/DeepSeek-V3-0324 - Highly capable reasoning model (64K context)
• deepseek-ai/DeepSeek-R1 - DeepSeek's logic-focused model featuring step-by-step logic (64K context)
• deepseek-ai/DeepSeek-R1-0528 - The most recent version of DeepSeek's reasoning model (64K context)
• meta-llama/Llama-3.1-8B-Instruct - Fast and highly capable 8B Llama model suited for general queries (128K context)

Configuration in Sypha

1. Open Sypha Settings: Click the gear icon (⚙️) to open settings in the Sypha panel.
2. Choose Provider: Find and select "Hugging Face" from the "API Provider" dropdown menu.
3. Enter Token: Paste the Hugging Face access token you created into the required field.
4. Select Your Model: Pick the model you want to experiment with from the "Model" dropdown list.

Tips and Notes

• Free Inference: The list of models above is currently running without cost through Hugging Face Inference Providers.
• Open Source: Every model is completely open-source, giving you the freedom to move to self-hosted versions whenever you want.
• Rate Limits: Be aware that the free tiers come with strict rate limits. You can find the exact details in the Hugging Face documentation.
• Model Availability: Server capacity at the inference providers dictates if a model is currently online or delayed, so availability can fluctuate.