Kimi K2.5 vs Kimi K2.6

The match-up at a glance

This comparison pits two open-weight models that you can install on your own servers against each other: Kimi K2.5 and Kimi K2.6, both built by Moonshot AI. The first dates from January 27, 2026, the second from April 20, 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.

What the benchmarks say

On our capability index, Kimi K2.5 scores 78.1 out of 100 and Kimi K2.6 84.4. Kimi K2.6 leads by 6.3 points.

On the 3 benchmarks reported for both, Kimi K2.5 wins 0 times and Kimi K2.6 3:

• GPQA Diamond (PhD-level science questions): Kimi K2.5 87.6%, Kimi K2.6 90.8%
• OTIS Mock AIME (competition mathematics problems): Kimi K2.5 92.2%, Kimi K2.6 96.1%
• SWE-bench Verified (fixing real software bugs): Kimi K2.5 73.8%, Kimi K2.6 76.7%

Pricing

Kimi K2.5 costs $0.58 per million input tokens and $3 per million output tokens; Kimi K2.6 costs $0.95 and $4.

For a typical agent task (read 20,000 tokens, write 2,000), expect $0.018 with Kimi K2.5 versus $0.027 with Kimi K2.6. Kimi K2.5 is 1.5 times cheaper, a gap that adds up quickly over 1,000 tasks: $9.4 difference.

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

Under the hood

Both models have the same context window: 262,144 tokens, roughly 393 pages.

Reasoning mode: Kimi K2.5 yes, Kimi K2.6 yes. Tool calling: Kimi K2.5 yes, Kimi K2.6 yes. Image input: Kimi K2.5 yes, Kimi K2.6 yes.

Both publish their weights: you can host them yourself.

Knowledge cutoff: January 2025 for Kimi K2.5, January 2025 for Kimi K2.6.

In short

Our reading of the data:

• Choose Kimi K2.5 for a lower cost.
• Choose Kimi K2.6 for a higher capability index and a more recent release.

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