kimi-k2-0905-preview vs zai-glm-4.6

Capability for the money

kimi-k2-0905-preview is 2.2× cheaper than zai-glm-4.6 and scores higher on following instructions (IFBench) and hard reasoning (GPQA Diamond) — there is no case for paying more here.

1. Following instructions

   extraction, classification, sticking to a format

   IFBench
   kimi-k2-0905-preview41.7%

   zai-glm-4.636.7%

   kimi-k2-0905-preview +5.0

2. Coding

   writing and fixing code

   LiveCodeBench
   kimi-k2-0905-preview61.0%

   zai-glm-4.656.1%

   kimi-k2-0905-preview +4.9

3. Tool use

   multi-step agent work against real tools

   τ²-bench
   kimi-k2-0905-preview73.4%

   zai-glm-4.676.9%

   zai-glm-4.6 +3.5

4. Hard reasoning

   graduate-level science and multi-step analysis

   GPQA Diamond
   kimi-k2-0905-preview76.7%

   zai-glm-4.663.2%

   kimi-k2-0905-preview +13.5

Averaged over the benchmarks both models sat, kimi-k2-0905-preview returns 58.8 points of pass rate per dollar of blended price against 24.5 — the ratio that decides whether the dearer model earns its rate.

Each figure is a pass rate: the share of that test set the model answered correctly, measured by a third party rather than self-reported by the lab. Rows are different tests of different difficulty, so compare a model against the other model, not one row against the next. And a benchmark is not your workload — it narrows the shortlist, it does not pick for you.