We Swapped Opus 4.6 for GLM 5.2 in a Side-by-Side Comparison

We ran GLM 5.2 in our Kilby AI harness. The cost was 7.4x cheaper than when we run Opus 4.6 inside Kilby.

Below is a comparison of Kilby’s earnings transcript sentiment analysis. We compared the production version of Kilby’s analysis, which sits on top of Claude Opus 4.6 with the test version, which swaps Opus 4.6 for GLM 5.2. The results were very similar, yet the cost differential was significant.

• The cost of the Kilby production analysis (Opus 4.6 model) was $0.26.

• The cost of the GLM 5.2 test run was $0.035. GLM was 7.4x cheaper to run.

See the comparison for yourself, we have included links to the PDF analysis:

• Kilby’s NVDA Transcript Analysis (production version using Claude Opus 4.6): HERE

• NVDA Transcript Analysis (Test version using GLM 5.2 model, which we are considering on-boarding to Kilby as a lower-cost model option): HERE

---