The models available now are one of the lowest ranking models for coding. Had my first session and was unimpressed by the solution provided. Maybe make the better models available for only coding/assement sessions (no need for high speed generations) or for higher teir customers. Something to thing about.
Coding Performance Rankings
The primary difference between these models lies in their architecture. Sol and Opus 5 utilize heavy internal reasoning steps, while Terra, Gemini 3.5 Flash, and Luna rely on direct, high-speed generation.
Sol (Rank 1: Elite Reasoning & Architecture)
Best for complex, multi-file software engineering.
Excels at debugging multi-step logic errors.
Generates highly optimized, production-ready code blocks.
Opus 5 (Rank 2: Advanced Coding & Logic)
Built for massive codebases and deep analysis.
Excellent at handling highly nuanced, custom edge cases.
Marginally slower than Sol but matches its structural accuracy.
Terra (Rank 3: Mid-Tier Balancing)
The go-to workhorse for day-to-day scripting.
Quickly writes standard APIs, tests, and boilerplates.
Struggles when handling highly abstract architectural decisions.
Gemini 3.5 Flash (Rank 4: High-Speed Refactoring)
Unmatched speed for quick code translation and reviews.
Ideal for ingesting large repositories to find context.
May output occasional syntax errors in complex logic. [1]
Luna (Rank 5: Simple Automation Only)
Limited to basic scripts and syntax corrections.
Prone to logic gaps in long programming tasks.
Best used strictly for high-volume, low-complexity formatting.