getlilac.com

What It Is

Lilac is a large-model inference API platform for developers, positioned as offering “frontier-speed inference, better prices.” Its core idea is to route inference requests to enterprise GPUs that are already powered on but underutilized, thereby reducing the per-token cost. It is not a chat app in the traditional sense, but rather model inference infrastructure for the backend of AI applications.

Core Capabilities and Integration

The page puts strong emphasis on being OpenAI-compatible: existing projects can continue using the OpenAI SDK and connect to Lilac simply by changing the base URL. It offers shared warm endpoints, claiming no cold starts, low latency, and high throughput. It also provides a real-time model status page showing metrics such as TPS, throughput, TTFT, and availability, updated roughly every 30 seconds from recent API traffic. Models currently listed include MiniMax M2.7, Kimi K2.6, GLM 5.1, and Gemma 4 31B, with context lengths of around 200K to 262K.

Pricing and Value

Lilac uses token-based billing, with no contracts, no commitments, and no minimum spend. Example pricing includes Gemma 4 31B at $0.11/M input and $0.35/M output; MiniMax M2.7 at $0.30/M input and $1.20/M output; Kimi K2.6 at $3.50/M output; and GLM 5.1 at $3.00/M output. For teams with fluctuating usage that want to avoid reserved instances or minimum commitments, the pricing model is relatively friendly.

Pros and Cons

The advantages are a low barrier to integration, transparent pricing, pay-as-you-go billing, warm endpoints that reduce cold starts, and support for some long-context models. The limitations are also clear: the crawled text does not disclose data privacy policies, log retention, SLA, payment methods, or enterprise support tiers. The model lineup is still limited, though the page says more models are coming soon. Because its performance depends on a routing mechanism for idle GPUs, production deployments should still run their own load tests.

Who It’s For and Access from China

Lilac is suitable for developers who need low-cost LLM APIs, AI SaaS teams, bulk text-generation businesses, and engineering teams looking to migrate OpenAI SDK-based applications to an alternative inference backend. The page does not state how well it works from China, and payment methods are not disclosed, so these should be treated as unknowns. Teams in mainland China may also want to evaluate local alternatives such as SiliconFlow, Alibaba Cloud Bailian, and Volcano Ark to reduce uncertainty around network connectivity and payments.