feat: refresh model list to best price-to-performance picks #36

Closed

Jc-965 wants to merge 0 commits from model-price-performance-update into main

Jc-965 commented

2026-08-06 22:50:02 +00:00

(Migrated from codeberg.org)

Summary

The model picker had fallen behind the current price-to-performance landscape. This update refreshes the list so each provider is represented by its strongest value option, with tool calling verified through OpenRouter since the agent needs to bind the CMU MCP tools.

GPT-5.6 Luna replaces GPT-5.4 Mini as the default. It performs better across every shared benchmark while costing roughly one-seventh as much.
DeepSeek V4 Flash, GLM 5.2, and MiniMax M3 have been added as strong, low-cost options for agentic workloads.
Claude Haiku 4.5 and Gemini 3.5 Flash Lite remain the Anthropic and Google options.
Users whose saved preference points to a removed model automatically fall back to the new default through the existing isValidModelId check, so no migration is required.

## Summary The model picker had fallen behind the current price-to-performance landscape. This update refreshes the list so each provider is represented by its strongest value option, with tool calling verified through OpenRouter since the agent needs to bind the CMU MCP tools. - GPT-5.6 Luna replaces GPT-5.4 Mini as the default. It performs better across every shared benchmark while costing roughly one-seventh as much. - DeepSeek V4 Flash, GLM 5.2, and MiniMax M3 have been added as strong, low-cost options for agentic workloads. - Claude Haiku 4.5 and Gemini 3.5 Flash Lite remain the Anthropic and Google options. - Users whose saved preference points to a removed model automatically fall back to the new default through the existing `isValidModelId` check, so no migration is required.