Moonshot AI · open weights
Kimi K2.6 API
Open-weight agentic model within a point of Claude Opus 4.6 on published coding benchmarks.
~40% below OpenRouter list for the same model. Prices verified August 14, 2026. Full market table on /pricing.
Call it
Model id: moonshotai/Kimi-K2.6 (case-sensitive). Any OpenAI SDK works.
from openai import OpenAI
client = OpenAI(
base_url="https://router.mingles.ai/v1",
api_key="sk-your-free-key",
)
resp = client.chat.completions.create(
model="moonshotai/Kimi-K2.6",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)Published benchmarks
| Benchmark | Kimi-K2.6 | Claude Opus 4.6 | GPT-5.4 |
|---|---|---|---|
| SWE-Bench Verified | 80.2 | 80.8 | — |
| SWE-Bench Pro | 58.6 | 53.4 | 57.7 |
| SWE-Bench Multilingual | 76.7 | 77.8 | — |
| Terminal-Bench 2.0 | 66.7 | 65.4 | 65.4 |
| LiveCodeBench v6 | 89.6 | 88.8 | — |
| BrowseComp | 83.2 | 83.7 | 82.7 |
| AIME 2026 | 96.4 | 96.7 | 99.2 |
| GPQA-Diamond | 90.5 | 91.3 | 92.8 |
Source: Moonshot AI — Kimi-K2.6 model card. Full comparison with sources on /compare/claude-api.
Specs
| Architecture | Mixture-of-experts, 1T total / 32B active parameters (per model card) |
| Strengths | Agentic coding, tool use, browsing — scores averaged over 10 runs |
| License | Open weights (see model card) |
On a plan
Plans hold a priced allowance, not a token count, so what it buys depends on the model. On Kimi K2.6 it works out to roughly:
On the coding table Moonshot publishes, K2.6 sits within a point of Claude Opus 4.6 (80.2 vs 80.8 on SWE-Bench Verified) and ahead on SWE-Bench Pro. It trails frontier models by a couple of points on hard reasoning (AIME, GPQA). Per blended token it costs about 1/15th of Opus — which is why "close on code" decides most agent workloads.
FAQ
▸How do I call Kimi K2.6 through the API?
POST /v1/chat/completions with "model": "moonshotai/Kimi-K2.6" — any OpenAI SDK works after changing base_url to https://router.mingles.ai/v1. Streaming and tool-calling included.
▸What does Kimi K2.6 cost?
$0.35 per 1M input tokens and $1.46 per 1M output (~40% below OpenRouter list). On a plan, the same money buys ~4–5× more: the $9 Mini plan covers roughly 39M Kimi tokens a month.
▸Is it as good as Claude?
On published coding benchmarks, within a point; on hard reasoning, a couple of points behind. Sources and the full table are on /compare/claude-api — run your own evals before switching production traffic.