Kimi K3 vs GLM-5.2
Kimi K3 and GLM-5.2 are the two leading open-weights models. K3 scores seven Intelligence Index points higher (60 vs 53) but costs twice as much hosted ($3/$15 vs $1.40/$4.40 per million tokens) and is far heavier to self-host: 2.8T total parameters against GLM's 753B. GLM-5.2 also generates over three times faster.
Side by side
| Kimi K3 | GLM-5.2 | |
|---|---|---|
| Maker | Moonshot AI | Z AI |
| Released | 2026-07-16 | 2026-06-16 |
| Context window | 1M tokens | 1M tokens |
| Modalities in | text, image | text |
| Input $/MTok | $3 | $1.4 |
| Output $/MTok | $15 | $4.4 |
| Open weights | Yes (Kimi K3 License (commercial terms apply)) | Yes (Open weights) |
Benchmark deltas
| Benchmark | Kimi K3 | GLM-5.2 | Δ |
|---|---|---|---|
| AA Intelligence Index | 60 | 53 | +7 |
Δ is Kimi K3 minus GLM-5.2, in points. Shaded cell wins the row. Dashed values are vendor-published, not yet reproduced — methodology.
Cost calculator
- Kimi K3
- $300/mo
- GLM-5.2
- $114/mo
- Difference
- $186/mo
List prices as of 2026-08-11; excludes caching, batch, and long-context tiers — see model pages for tier notes.
Verdict
K3 is the capability champion of open weights; GLM-5.2 is the operability champion. Seven Intelligence Index points is a real gap — K3 sits with the closed frontier while GLM plays a tier below — but GLM-5.2's 40B active parameters, 142 tokens-per-second serving speed, and friendlier footprint make it the model more teams can actually run. Hosted, GLM costs under a third as much. Choose K3 when peak reasoning justifies the bill and the hardware; choose GLM-5.2 for self-hosted production at scale.
Verdicts are analysis, not measurement, and are labeled as such per our editorial policy.
FAQ
- Is Kimi K3 cheaper than GLM-5.2?
- No. GLM-5.2 is cheaper: $1.4 input / $4.4 output per million tokens against $3 / $15 for Kimi K3 (prices as of August 11, 2026).
- Which has the longer context window — Kimi K3 or GLM-5.2?
- Kimi K3 does — 1M tokens against 1M for GLM-5.2.
- Which is the more capable model, Kimi K3 or GLM-5.2?
- Kimi K3 scores higher on the Artificial Analysis Intelligence Index: 60 (measured August 11, 2026) against 53 for GLM-5.2 (measured August 11, 2026).
Related news
Comparison published August 11, 2026; underlying data last changed August 11, 2026.