GLM-5.2 (on Mistral) vs Kimi K3
Both are Flagship-tier models, from Mistral AI and Moonshot AI respectively. GLM-5.2 (on Mistral) is the cheaper choice overall at a 4:1 input:output ratio.
Which is cheaper
- Input price: GLM-5.2 (on Mistral) is cheaper ($1.40 vs $3.00).
- Output price: GLM-5.2 (on Mistral) is cheaper ($4.40 vs $15.00).
- Blended, 4:1 input:output ratio: GLM-5.2 (on Mistral) is 63% cheaper ($2.00 vs $5.40 per weighted 1M tokens).
Side by side
| Spec | GLM-5.2 (on Mistral) | Kimi K3 |
|---|---|---|
| Provider | Mistral AI | Moonshot AI |
| Tier | Flagship | Flagship |
| Context window | 1M | 1.05M |
| Max output | 128K | — |
| Input $/1M | $1.40 | $3.00 |
| Output $/1M | $4.40 | $15.00 |
| Cache read $/1M | $0.14 | $0.30 |
| Modalities | text | text |
GLM-5.2 (on Mistral). A Z.ai model resold by Mistral and served on Mistral's EU infrastructure, listed as zai-glm-5-2 on Mistral's own pricing page. Cache read is $0.14 per 1M.
Kimi K3. input is cache-miss (standard) price per 1M tokens; cache_read is cache-hit price ($0.30).
Sources
GLM-5.2 (on Mistral) — verified August 29, 2026
Kimi K3 — verified August 29, 2026
Full GLM-5.2 (on Mistral) pricing page → · Full Kimi K3 pricing page →