Model comparison
Kimi K2.5 vs Kimi K2 Thinking
Price history, scores and specs for both models, line by line. Every value shows where it came from and when we last confirmed it.
Moonshotai
Kimi K2.5
Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm.
Moonshotai
Kimi K2 Thinking
Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning.
Quick take
Where each one leads
- Kimi K2.5 costs less to send text to: $0.532 per million tokens, against $0.600 per million tokens for Kimi K2 Thinking.
- Kimi K2 Thinking costs less per word it writes: $2.50 per million tokens, against $3.33 per million tokens for Kimi K2.5.
- Kimi K2.5 scores higher on the Artificial Analysis Intelligence Index: 23.5 to 22.
- Kimi K2.5 scores higher on the Artificial Analysis Coding Index: 46.8 to 21.
- More people run Kimi K2.5 through OpenRouter than Kimi K2 Thinking right now.
Each line comes from a checked value in the tables below. Pick on your workload: price and usage matter for high volume, context and scores for long documents and agent loops.
Price history
Prices side by side, day by day
- Kimi K2.5
- Kimi K2 Thinking
- Kimi K2.5
- Kimi K2 Thinking
Prices are the listed OpenRouter base rates in US dollars. A day we could not confirm is left out of the line, never filled in.
Price
What each one costs now
| Measure | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Input price | $0.532 per million tokens Leads | $0.600 per million tokens |
| Output price | $3.33 per million tokens | $2.50 per million tokens Leads |
Scores
Benchmarks, with source and date
| Measure | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Artificial Analysis Intelligence Index | 23.5 Leads | 22 |
| Artificial Analysis Coding Index | 46.8 Leads | 21 |
| LMArena text rating (Elo) | 1,445 | Not checked yet |
| LMArena web development rating (Elo) | 1,436 | Not checked yet |
| LMArena vision rating (Elo) | 1,269 | Not checked yet |
| Design Arena rating (Elo) | 1,245 Leads | 1,109 |
Specs
Context, modalities and tool use
| Measure | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Context window | 262,144 tokens | 262,144 tokens |
| Reads | Text, Images | Text |
| Writes | Text | Text |
| Tool calling | Yes | Yes |
| Reasoning mode | Yes | Yes |
| Open weights | Yes | Yes |
| Listed since | 27 Jan 2026 | 6 Nov 2025 |
Usage
How much each one is used
| Measure | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Tokens served on OpenRouter in a day | 3.8B Leads (for 10 Nov 2026) | 338.8M (for 10 Nov 2026) |
| Requests on OpenRouter in a day | 261.6K Leads (for 10 Nov 2026) | 17K (for 10 Nov 2026) |
| Tool calls on OpenRouter in a day | 46.7K Leads (for 10 Nov 2026) | 4.4K (for 10 Nov 2026) |
| Token usage, last 7 days | Up 5% on the week before Leads (for 10 Nov 2026) | Down 42% on the week before (for 10 Nov 2026) |
- Kimi K2.5
- Kimi K2 Thinking
How this works
One link, either order
This link stays the same whichever order you name the two models in, so you can share it as it is. A line appears only when at least one model has a value we confirmed from a source; values we have not confirmed are left out, never guessed. To add a third record of any type, use the open comparison. Read how we check facts.