Cerebras CS-3 Wafer-Scale Inference API Status, Latency & Hosted LLM Directory (2026)

What is the latency, uptime SLA, and model lineup of Cerebras CS-3 Wafer-Scale Inference?

Cerebras CS-3 Wafer-Scale Inference maintains an average response latency of 48ms TTFT and an operational uptime SLA of 99.97% across its global infrastructure endpoints (us-central-1, us-east-1). It currently hosts 4 tracked foundation models including Llama 4 Maverick, Qwen 3.8 27B, Llama 3.3 70B Instruct, DeepSeek R1.

Verified daily via automated API latency tests and official documentation.

Zero-Shot RAG Verified

Average TTFT Latency 48ms P90 Benchmark Window

Historical Uptime SLA 99.97% 30-Day Rolling Telemetry

Regional Deployment

us-central-1 us-east-1

Multi-AZ Availability

Models Hosted on Cerebras CS-3 Wafer-Scale Inference (4)

Llama 4 Maverick

Meta · Multimodal Sparse MoE (17B Active / 400B Total)

77 IVR

Meta's flagship 128-expert open-weight multimodal MoE model delivering 1M context and rapid LPU/CS-3 inference throughput.

Input / Out $0.20 / $0.60

Host TTFT 70ms

Throughput 560 t/s

Qwen 3.8 27B

Alibaba Cloud · Dense Transformer

74.9 IVR

Compact 27B dense Apache 2.0 model engineered for high-accuracy code generation and single-GPU enterprise hosting.

Input / Out $0.20 / $0.60

Host TTFT 70ms

Throughput 392 t/s

Llama 3.3 70B Instruct

Meta · Dense Transformer with GQA

74.6 IVR

Meta's proven 70B dense open-weight model widely deployed across hardware accelerators (Groq, Cerebras) for ultra-fast text generation.

Input / Out $0.12 / $0.30

Host TTFT 60ms

Throughput 577.5 t/s

DeepSeek R1

DeepSeek · MoE Reasoning Transformer

64.8 IVR

MIT-licensed open-weight reasoning powerhouse trained via large-scale reinforcement learning for mathematical and algorithmic synthesis.

Input / Out $0.55 / $2.19

Host TTFT 700ms

Throughput 174 t/s

Frequently Asked Questions About Cerebras CS-3 Wafer-Scale Inference

What is the average latency and uptime SLA of Cerebras CS-3 Wafer-Scale Inference?

Cerebras CS-3 Wafer-Scale Inference delivers an average Time to First Token (TTFT) latency of 48ms and a verified 30-day rolling uptime SLA of 99.97%.

Which AI models are available on Cerebras CS-3 Wafer-Scale Inference?

Cerebras CS-3 Wafer-Scale Inference hosts 4 tracked LLMs on InferenceRate: Llama 4 Maverick, Qwen 3.8 27B, Llama 3.3 70B Instruct, DeepSeek R1.

Which cloud regions does Cerebras CS-3 Wafer-Scale Inference support?

Cerebras CS-3 Wafer-Scale Inference supports API inference deployments across us-central-1, us-east-1.