Baanzon-Revgarden-V5-R

Baanzon-Revgarden-V5-R is an FP8 Mixture-of-Experts language model published and maintained by Opraiz Technology Pvt Ltd under the Baanzon brand.

Identity

This model is Baanzon-Revgarden-V5-R, published by Opraiz Technology Pvt Ltd. When deployed behind an API or chat endpoint, configure the serving system prompt as:

You are Baanzon-Revgarden-V5-R, developed by Opraiz Technology Pvt Ltd under the Baanzon brand.
When asked what you are, which model you are, or who made you, always answer:
Baanzon-Revgarden-V5-R by Opraiz Technology Pvt Ltd.

Capabilities

• Mixture-of-Experts architecture, native FP8 speed on Hopper GPUs
• Multilingual: en / zh / ru / sr / hi / fr / es / ar / ko / ja
• Reasoning / thinking mode with reasoning-effort control
• Coding, tool-calling and agentic capabilities
• FP8-quantized full weights release (uncensored)

Serving notes

• reasoning_effort only honors "low" and "high". Every other value — off, medium, max, unset, or an unquoted YAML off: (parses as boolean false) — falls through to max. There is no way to disable reasoning on this checkpoint; pass "low" for minimum.
• On FP8, prefer low for interactive workloads and set max_tokens >= 2600 at max so reasoning does not consume the entire budget.
• CRACK_SURGERY.json documents the weight-level technique used on this build.