Kimi K2.5vsKimi K2.6
Both language models, compared at each maker’s own API list prices. Every figure links to its source.
Compare at your usage
Every model priced for the same request8,000 tokens in, 2,000 out
| Kimi K2.5 on Alibaba Cloud Model Studio | Kimi K2.6 on Moonshot AI Platform | |
|---|---|---|
| Rate applied | Base rate | Base rate |
| Inputper 1M tokens | $0.574 | $0.95 |
| Cache readper 1M tokens | $0.115 | $0.16 |
| Outputper 1M tokens | $3.011 | $4.00 |
| Cost per request | $0.01061 | $0.0156 |
| Per 1,000 requests | $10.61 | $15.60 |
- Kimi K2.5 Alibaba Cloud Model Studio pricing · as of Nov 2, 2026 · kimi-k2.5 model · as of Nov 1, 2026
- Kimi K2.6 Moonshot AI platform pricing · as of Nov 1, 2026
Every model is priced for the same request at its maker’s own list prices (each with its source and date), so bold marks the cheapest. A model that can’t serve the request is named, not ranked; a condition its seller doesn’t price says which rate was used.
Cost per request as prompts grow
2K tokens out, list price on each maker’s API. A line ends at the model’s context window.
- Kimi K2.5 · Alibaba Cloud Model Studio
- Kimi K2.6 · Moonshot AI Platform
Where each stands among all scored models
Each line runs from the lowest to the highest score on the board; the tick is the median.
- Kimi K2.5
- Kimi K2.6
Key differences
These differences use verified vendor specifications or each maker’s official API on the same region, currency, meter and billing conditions. Different conditions remain side by side in the full table; they do not establish which model is cheaper. Declared features describe support, not measured quality or a recommendation. Missing evidence does not mean a feature is unsupported.
Full comparison
| Kimi K2.5 Moonshot AI | Kimi K2.6 Moonshot AI | |
|---|---|---|
| Overview | ||
| Released | Jan 2026 | Apr 2026 |
| Open weights | ||
| Knowledge cutoff | Jan 2025 | Jan 2025 |
| Specs | ||
| Context | 256K | 262.14K |
| Max output | 16.38K | 262.14K |
| Input modalities | Text, Image, Video | Text, Image, Video |
| Output modalities | Text | Text |
| Reasoning | ||
| Reasoning can be off | ||
| Max thinking budget | 81.92K | Not stated |
| Tool use | ||
| Structured output | — | |
| Pricing · per 1M tokens | ||
| Input price | $0.574 | $0.95 |
| Output price | $3.011 | $4.00 |
| Cache read price | $0.115 | $0.16 |
| Cache write price | $0.718Explicit cache | — |
| Blended price (3 input : 1 output) | $1.18325 | $1.7125 |
| Cited scores | ||
| Epoch Capabilities Index | 148.7 | 151.6 |
| LMArena Text | — | 1,461 |
| LMArena Text · Coding | — | 1,516 |
| LMArena Vision | — | 1,282 |
| LMArena Text · Chinese | — | 1,517 |
| Epoch AI · GPQA Diamond | — | 90.8% |
| Epoch AI · FrontierMath Tiers 1–3 | — | 57.2% |
| Epoch AI · SWE-bench Verified | 73.8% | 76.7% |
Headline prices may come from different channels, regions or tiers. Check each source before choosing a model; missing facts stay blank.
Model details and sources
- Kimi K2.5 → Available
- Kimi K2.6 → Deprecated · as of Nov 1, 2026 Newer models in this family (our classification): Kimi K3
Cited scores
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Nov 5, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Nov 4, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Oct 7, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · EQ-Bench: Longform Writing · MIT · as of Oct 28, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Oct 28, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · EQ-Bench: 4 · MIT · as of Oct 24, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Sep 13, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Oct 28, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Oct 19, 2026