Can Open Source Models Beat Opus at a Fraction of the Cost?

I put five open-source models (Kimi K2.6, MiniMax M2.7, GLM 5.1, DeepSeek V4 Pro, and Qwen 27B) head-to-head against Opus using the Copilot CLI. The question: can you actually replace Opus with a nearly-free open-source model and save a TON of money? How to add open-source models to Copilot CLI:    • Yes, you can use open source models with C...   PRD used in the test: https://gist.github.com/burkeholland/... 0:00 Intro 0:48 The test app 2:05 How we're running the tests 3:14 The contenders and what they cost 6:32 How we're scoring 7:47 Baseline: Claude Opus 4.6 11:13 Kimi K2.6 13:19 MiniMax M2.7 15:31 GLM 5.1 18:12 DeepSeek V4 Pro 20:22 Qwen 27B 23:29 The verdict #ai #llm #coding