Please Add MiniMax-M2.5 (SOTA Coding at Ultra-Low Cost)

I am requesting the addition of MiniMax-M2.5 to the Venice.AI model lineup. This model represents a breakthrough in high-intelligence, low-cost inference that aligns perfectly with Venice’s goal of providing powerful, uncensored AI.

Why MiniMax-M2.5?

1. Top-Tier Performance: It currently scores 80.2% on SWE-bench Verified, effectively matching Claude 4.5 and GPT-5 class models in coding and complex reasoning tasks.

2. Unmatched Cost Efficiency: * M2.5 is roughly 1/10th the cost of Gemini 3 Pro and GPT-5 for similar outputs.

   • Current API rates are as low as $0.30 per 1M input tokens, making it one of the most economical "Large" class models ever released.

   • Its Mixture-of-Experts (MoE) architecture (only 10B active parameters) allows for extremely fast inference speeds (~100 TPS), which would reduce latency for Venice Pro and API users.

3. Open Weights: The weights are available on Hugging Face, allowing Venice to host the model directly to ensure the privacy and lack of censorship that the community expects.

4. Large Context: With a 200k context window, it is ideal for the deep research and document analysis features Venice is known for.

Impact on Venice Users: Adding M2.5 would allow Venice to offer a "Pro" level coding and reasoning experience while keeping overhead low. This could potentially enable more generous limits for Free users or higher-velocity agents for Pro subscribers.

Source: https://huggingface.co/MiniMaxAI/MiniMax-M2.5