GPT‑OSS‑class large model (for example gpt‑oss‑120B)

Chat / reasoning / vision

Large context (check Cerebras / Orq docs for current limits)

Complex reasoning, analysis, coding, and high‑stakes tasks where you want a large open model served at very high throughput on Cerebras Cloud

Premium – Cerebras Inference higher tiers; typically positioned above smaller models, with pay‑per‑token or pay‑per‑model options

Mid‑sized open model (for example Llama 3.x 8B–70B, Qwen 2.5 / 3 mid‑tier)

Chat / reasoning / coding

Medium–large context (verify in Cerebras / Orq docs)

Everyday production workloads, RAG, coding, product features, and workflows that balance quality with speed and cost

Mid‑tier – Cerebras Developer / paid plans; often in the ~0.10–0.60 USD per million token range depending on model choice

Smaller / fast open model (for example Llama 3.1 8B, lighter GPT‑OSS tiers)

Chat / fast / cost‑efficient

Medium context (confirm in docs)

Fast, lower‑cost tasks such as classification, extraction, lightweight chat, routing, and high‑volume support workflows

Cost‑efficient – Cerebras Developer tier pricing; lower per‑million token cost designed for high‑throughput workloads and experimentation

Chat

Cerebras‑served chat models (for example GPT‑OSS, Llama, Qwen on Cerebras)

Call Cerebras models directly via the Cerebras Inference API for chat, reasoning, coding, and multimodal tasks

Use Cerebras models through Orq.ai’s OpenAI‑compatible endpoint, adding routing, tracing, evals, budgets, and governance controls around each request

Code

Open models on Cerebras tuned or suitable for coding/analysis

Use Cerebras directly for code generation, debugging, refactoring, and agentic coding workflows, often with higher tokens‑per‑second vs GPU clouds

Route coding workloads through Orq.ai, compare Cerebras‑served models against other providers, and monitor cost, latency, and quality from one control layer

Embeddings

Embedding providers configured in Orq.ai

Cerebras focuses on high‑performance serving of LLMs; embeddings may be handled by separate models or providers

Use Orq.ai to route embedding workloads to supported embedding providers while keeping Cerebras for reasoning, generation, and agent steps

Model rates

Cerebras model pricing may differ depending on whether you:

* connect your own Cerebras Cloud account and API key (BYOK), or

* use Cerebras‑hosted models billed through Orq.ai where available Check the Orq.ai pricing page and your workspace’s provider configuration for current per‑model rates, quotas, and plan details.