AI.cc adds 500+ Hugging Face models to its unified API

AI.cc said enterprise customers can now access more than 500 open-source models from Hugging Face through its existing OpenAI-compatible API, with no self-hosting or new integration required. The move expands the Singapore platform’s catalog to 800+ models and is aimed at lowering enterprise AI costs and operational complexity.

Why it matters: - AI.cc is trying to remove the biggest barrier to enterprise use of open-source AI: the need to run and maintain models in-house. - The company says the new setup can cut blended token costs by 40% to 65% versus proprietary-only deployments. - Enterprises can use open-source models for routine work and reserve frontier models for higher-stakes tasks.

What happened: - AI.cc announced on May 30, 2026, that enterprise customers can access 500+ open-source models from the Hugging Face Hub through AI.cc’s unified API. - The Singapore-based platform said the new catalog is available through its existing OpenAI-compatible endpoint. - The expanded catalog brings AI.cc to 800+ total models across proprietary and open-source categories. - Customers do not need a new SDK, a separate Hugging Face Inference API account, or self-hosting infrastructure.

The details: - The expanded catalog includes Llama 4, Mistral Large 3, GLM-5.1, DeepSeek V4, Gemma 4, and hundreds of additional community models. - AI.cc said the open-source catalog was selected based on download volume, benchmark performance, license permissiveness, and enterprise deployment suitability. - The list spans foundation and reasoning models, specialized fine-tuned models, and embedding and retrieval models. - Foundation models include the full Llama 4 family, Mistral Large 3, Mistral Small 4, Devstral 2, the full Qwen 3.x series, Gemma 4, GLM-5.1, GLM-5V-Turbo, DeepSeek V4, Falcon 3, Falcon 2, and Arcee Trinity. - Specialized models cover biomedical, legal, financial, multilingual, and code-focused use cases. - The catalog also includes models tuned for regulatory compliance, clinical documentation, financial report generation, and Southeast Asian language coverage. - AI.cc said the open-source models are accessible through the same API key and call structure used for proprietary models such as Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro. - Code examples show only a model-name change is needed to switch from proprietary to open-source models. - Token usage across proprietary and open-source models is consolidated in AI.cc’s billing dashboard. - The OpenClaw agent framework can route tasks across model types inside one workflow. - AI.cc said a single agent workflow can assign GLM-5.1 to coding subtasks, Llama 4 Scout to long-context retrieval, Mistral Small 4 to classification, and Claude Opus 4.7 to high-stakes reasoning. - AI.cc said a representative document-processing workflow could route document ingestion and OCR to Gemma 4 12B, extraction to Qwen 3.5 9B, summarization to Mistral Large 3, risk analysis to Claude Opus 4.7, and output formatting to GLM-5.1. - The company said that kind of routing can push about 70% of token volume to open-source models priced below $0.50 per million input tokens. - AI.cc said the resulting blended cost could fall to $0.35 to $0.65 per million tokens, versus $5.00 to $18.00 per million tokens for workflows routed entirely through frontier proprietary models. - Free-tier access includes evaluation quota for all 500+ open-source models. - Enterprise customers can buy dedicated endpoints with SLA guarantees for specific models. - The complete catalog, pricing table, benchmark data, and documentation are posted at the open-source model catalog. - Enterprise inquiries are directed to AI.cc enterprise plans.

Between the lines: - The announcement reflects a broader shift in 2026: open-source models are now close enough to proprietary systems on many enterprise tasks to justify production use. - AI.cc is positioning unified routing, not model ownership, as the core enterprise advantage. - Licensing has also become less of a blocker, with many of the listed models now carrying commercial-friendly terms. - The company is betting that procurement teams care as much about cost control and operational simplicity as they do about model quality.

What's next: - Enterprises can start testing the expanded catalog immediately through AI.cc’s current API setup. - AI.cc is steering customers toward tiered routing architectures that mix open-source and proprietary models based on task complexity. - Dedicated inference capacity and SLA-backed access remain available for teams that need guaranteed performance.

The bottom line: - AI.cc is turning Hugging Face’s open-source catalog into a plug-in option for enterprise teams that want lower costs without rebuilding their AI stack.