Phase 2 - Model selection

[arekay-nv]

Would like to agree on a small list of models that we would like to support for Q2 rollout. This will help drive rules/dataset/constraint discussion as well as allow us to do mock submissions to foresee any issues.
Current proposal include

• deepseek-ai/DeepSeek-R1 (large MoE)
• openai/gpt-oss-120b (medium MoE)
• Qwen/Qwen3.5-9B (small dense, newer) or meta-llama/Meta-Llama-3-8B-Instruct (small dense, older)

[attafosu]

I'd recommend to keep the Llama3.1-8B in the suite. While old, it's still actively being deployed in datacenter applications. It's already enabled and should be much more familiar for new submitters to the endpoints.