GLM-5.3 versus Kimi K2.5

What are we comparing?

GLM-5.3 and Kimi K2.5 are two open-weight models that you can install on your own servers, one from Zhipu AI (Z.ai), the other from Moonshot AI. GLM-5.3 was released on August 14, 2026, Kimi K2.5 on January 27, 2026.

All figures below come from public sources checked on October 22, 2026: models.dev and OpenRouter for prices and capabilities, Epoch AI for scores.

Which one is more capable?

On our capability index, GLM-5.3 scores 78.8 out of 100 and Kimi K2.5 78.1. The gap is too small to separate them.

On the 2 benchmarks reported for both, GLM-5.3 wins 1 time and Kimi K2.5 1:

• GPQA Diamond (PhD-level science questions): GLM-5.3 90.9%, Kimi K2.5 87.6%
• OTIS Mock AIME (competition mathematics problems): GLM-5.3 91.1%, Kimi K2.5 92.2%

Pricing

GLM-5.3 costs $1.2 per million input tokens and $4.18 per million output tokens; Kimi K2.5 costs $0.58 and $3.

For a typical agent task (read 20,000 tokens, write 2,000), expect $0.032 with GLM-5.3 versus $0.018 with Kimi K2.5. Kimi K2.5 is 1.8 times cheaper, a gap that adds up quickly over 1,000 tasks: $14.67 difference.

Run the numbers with your own volumes in the agent cost calculator.

Under the hood

GLM-5.3 reads more in one go: 1,000,000 tokens of context (roughly 1,500 pages), versus 262,144 for Kimi K2.5.

Reasoning mode: GLM-5.3 yes, Kimi K2.5 yes. Tool calling: GLM-5.3 yes, Kimi K2.5 yes. Image input: GLM-5.3 no, Kimi K2.5 yes.

Both publish their weights: you can host them yourself.

In short

Our reading of the data:

• Choose GLM-5.3 for a larger context window and a more recent release.
• Choose Kimi K2.5 for a lower cost and image input.

This verdict relies on public figures; it does not replace a trial on your own tasks.