What model sits between Qwen 3.8 27b and Flash next for coding?

A user is seeking a coding model that balances performance and hardware requirements, specifically positioned between Qwen 3.8 27b and Flash next. They desire a decode speed of 75-100 on an RTX 5090 with 96GB RAM for efficient iterations. The user is looking for a model that offers improved coding capabilities over Flash next without the decode speed sacrifice of Qwen 3.8 27b, or they might use Flash next for planning and 27b for implementation.

This user's specific hardware setup and decode speed requirements for a coding model offer a concrete comparison point unlike general performance benchmarks.