// head_to_head
Kimi K2.5 vs GLM 5.2
Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.
Kimi K2.5 is the cheaper of the two; neither can be ranked on quality here.
Kimi K2.5 does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
moonshotai
Kimi K2.5
- Blended / 1M
- $0.900
- Context
- 262K
- Released
- Jan 27, 2026
- Overall score
- Not evaluated
z-ai
GLM 5.2
- Blended / 1M
- $0.998
- Context
- 1.0M
- Released
- Jun 16, 2026
- Overall score
- 73.2
Specs and pricing
| Metric | Kimi K2.5 | GLM 5.2 |
|---|---|---|
| LiveBench overall Mean of the seven LiveBench category scores, 0–100. Higher is better. | — | 73.2 |
| Cost per point Measured benchmark spend divided by overall score — dollars per point of capability. | — | $0.1260 |
| Blended price / 1M 3:1 input:output mix, the usual shape of production traffic. | $0.900win | $0.998 |
| Input price / 1M | $0.450win | $0.650 |
| Output price / 1M | $2.25 | $2.04win |
| Cached input / 1M Price of an input token served from the prompt cache, where the provider publishes one. | $0.070win | $0.121 |
| Context window | 262K | 1.0Mwin |
| Max output tokens | 236Kwin | 131K |
Benchmarks by category
An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Kimi K2.5 on top, GLM 5.2 below, both out of 100.
What each one costs to run
Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.
| Workload | Kimi K2.5 | GLM 5.2 |
|---|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $260.64/mo | $281.15/mo |
| RAG assistant 8K in / 600 out × 100K requests | $343.00/mo | $430.59/mo |
| Coding agent 40K in / 4K out × 20K requests | $327.20/mo | $386.79/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $599.75/mo | $776.27/mo |
| Bulk classification 500 in / 20 out × 5M requests | $1,160/mo | $1,564/mo |
Which should you pick?
You need to fit large documents in one call
GLM 5.2
Wider context window — 1.0M against 262K.
You are cost-constrained
Kimi K2.5
Cheaper on blended list price at $0.900 per million tokens.
Kimi K2.5 vs GLM 5.2 FAQ
Which is better, Kimi K2.5 or GLM 5.2?
Kimi K2.5 is the cheaper of the two; neither can be ranked on quality here. Kimi K2.5 does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
Is Kimi K2.5 cheaper than GLM 5.2?
Kimi K2.5 is cheaper. On a 3:1 input:output blend, Kimi K2.5 lists at $0.900 per million tokens and GLM 5.2 at $0.998 — Kimi K2.5 is 11% cheaper. Input and output are priced separately — Kimi K2.5 charges $0.450 in and $2.25 out, GLM 5.2 charges $0.650 and $2.04 — so the model that looks cheaper flips depending on how output-heavy your workload is.
Does Kimi K2.5 or GLM 5.2 have a bigger context window?
GLM 5.2 has the larger context window: 262K for Kimi K2.5 against 1.0M for GLM 5.2. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.
Do Kimi K2.5 and GLM 5.2 support prompt caching?
Both publish a cached-input rate: $0.070 per million for Kimi K2.5 and $0.121 for GLM 5.2, against full input rates of $0.450 and $0.650. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.
Related comparisons
How these numbers are produced
- Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
- Scores — LiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
- Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
- “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.
Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.