Cerebras

Cerebras delivers ultra-fast AI inference (over 1,000–2,000+ tokens/second) powered by their CS-3 Wafer-Scale Engine (WSE-3), making coding agent pair-programming interactions and large repository scans virtually instantaneous.

Website: https://cloud.cerebras.ai/

---

Getting an API Key

1. Go to Cerebras Cloud.
2. Sign in or create an account.
3. Navigate to API Keys and generate a new key (csk-...).
4. Store your key securely.

---

Supported Models

Cerebras provides ultra-high-speed inference for top open-weights models:

High-Speed Coding & Workhorse Models

• llama-3.3-70b - (Recommended) Meta's flagship 70B foundation model with full 128k context running at 1,000+ tokens/sec
• llama3.1-70b - Versatile high-speed model for code refactoring and tool calling
• llama3.1-8b - Ultra-low-latency model (2,000+ tokens/sec) for instant completions and rapid script generation
• qwen-2.5-coder-32b - Dedicated open coding model with high-speed generation

Deep Reasoning Models

• deepseek-r1-distill-llama-70b - Distilled chain-of-thought reasoning model powered by Cerebras wafer-scale acceleration

---

Configuration in PostQode

1. Open PostQode Settings ().
2. Select Cerebras in the API Provider dropdown.
3. Enter your Cerebras API Key (csk-...) from cloud.cerebras.ai.
4. Select your model from the dropdown (e.g. llama-3.3-70b or deepseek-r1-distill-llama-70b).
5. Click Save Settings to connect.

---

Recommended Usage

• Ultra-Fast Subagent Exploration: Ideal for pairing with PostQode subagents (/subagent) where multiple parallel queries need immediate turnarounds.
• Rapid Planning & Prototyping: Instant token streaming enables interactive reasoning without waiting for conventional GPU queue delays.