SambaNova Cloud

SambaNova Cloud delivers full-precision (16-bit) AI inference at blazing speeds (up to 1,000+ tokens/second) powered by custom SambaNova Reconfigurable Dataflow Units (SN40L RDUs). This guarantees zero accuracy loss from heavy quantization while sustaining high throughput for agentic coding.

Website: https://cloud.sambanova.ai/

---

Getting an API Key

1. Navigate to SambaNova Cloud.
2. Create an account or sign in.
3. Head to the API Keys section in your dashboard and generate a new key.
4. Copy and store your key safely.

---

Supported Models

SambaNova serves full 16-bit precision models with high throughput and 128k context windows:

Coding & Agentic Execution

• Qwen2.5-Coder-32B-Instruct - (Recommended for Coding) Surgical diffs, unit testing, and multi-file code editing
• Meta-Llama-3.3-70B-Instruct - (Recommended for General Development) Flagship 70B open-weights model for architecture and coding
• Meta-Llama-3.1-8B-Instruct - Low-latency model for rapid commit generation and simple edits

Deep Reasoning & Planning

• DeepSeek-R1 - (Recommended for Planning) Frontier 671B chain-of-thought reasoning without quantization degradation
• DeepSeek-R1-Distill-Llama-70B - High-speed distilled reasoning model

Ultra-Scale Foundation

• Meta-Llama-3.1-405B-Instruct - Full 405B parameter model served at 16-bit precision for high-complexity architectural analysis

---

Configuration in PostQode

1. Open PostQode Settings ().
2. Select SambaNova in the API Provider dropdown.
3. Enter your SambaNova API Key from cloud.sambanova.ai.
4. Select your model from the dropdown (e.g. Meta-Llama-3.3-70B-Instruct, Qwen2.5-Coder-32B-Instruct, or DeepSeek-R1).
5. Click Save Settings to connect.

---

Key Advantages

• 16-bit Full Precision: Unlike heavily quantized 4-bit/8-bit hosting, SambaNova preserves mathematical accuracy and code indentation integrity.
• Sustained 1,000+ Tok/s: Ideal for massive codebase scanning and high-throughput subagent pipelines.