← Model list

Kimi K2 Thinking

moonshotai·moonshotai/kimi-k2-thinking·Deprecated·Open weights·kimi-thinking series

Thinking Kimi model for slower research passes, planning, and hard technical questions

At a glance
Official price$0.60 / $2.50 per 1M
Cache read $0.15 · Moonshot AI
Lowest paid$0.47 / $2.00
Vercel AI Gateway Cloud · 1.5× spread
Context262,144
⚠ Providers report 32,768–262,144; the table below is authoritative
Output limit262,144
Capabilities
ReasoningTool useStructured outputTemperatureAttachments
⚠ Providers report capability flags inconsistently
Modalities
Text
Knowledge cutoff2024-08
Released / updated2025-11-06 / 2025-11-06

Quality & performanceArtificial Analysis · Intelligence Index v4.1

Intelligence33.5
Value39
Coding
Agentic
Output speed121 tok/s
TTFT1.24 s
Value formulaIQ 33.5 ÷ min blended $0.853 = 39

Quality is independently evaluated by Artificial Analysis. Speed/latency are model-level medians.

Available at 21 providers19 with public prices

ProviderTierInputOutputCache readCache writeContextOutput limitStatus
Vercel AI GatewayCloud$0.47$2.00$0.141216,144216,144
HeliconeGateway$0.48$2.00256,000262,144
OpenCode ZenGateway$0.40$2.50$0.40262,144262,144deprecated
IO.NET
moonshotai/Kimi-K2-Thinking
Cloud$0.55$2.25$0.275$1.1032,7684,096
302.AIGateway$0.575$2.30262,144262,144
NanoGPTGateway$0.60$2.50$0.15262,14498,304
Vertex
moonshotai/kimi-k2-thinking-maas
First-party$0.60$2.50$0.06262,144262,144deprecated
Hugging Face
moonshotai/Kimi-K2-Thinking
Cloud$0.60$2.50$0.15262,144262,144
OpenRouterGateway$0.60$2.50$0.15262,144100,352
Moonshot AIOfficialFirst-party$0.60$2.50$0.15262,144262,144

Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.

Your usage cost

1Vercel AI Gateway$154.52
2Vertex$180.20
3DevPass (LLM Gateway)$180.20
4IO.NET$189.50
5NanoGPT$191.00
6Hugging Face$191.00
8Moonshot AI · Official$191.00
Switch to Vercel AI Gateway to save $36.48/mo (19%). The gap is small, so staying on the official channel is fine.

Benchmark1 items

NameConditionsScoreMetricSource
SWE-Bench Verified71.3resolvedSource ↗

The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.

Reasoning control

Toggle (on / off)budget_tokens ≥ 1 ≤ 32,768effort = high

16 / 21 providers expose no reasoning control (reasoning_options: []).

Related models

Price historyone sample accumulated per data sync

Input listOutput listMin blended
$2.50$02026-08-052026-08-132026-08-05 · Input list $0.602026-08-06 · Input list $0.602026-08-07 · Input list $0.602026-08-08 · Input list $0.602026-08-09 · Input list $0.602026-08-10 · Input list $0.602026-08-11 · Input list $0.602026-08-12 · Input list $0.602026-08-13 · Input list $0.602026-08-05 · Output list $2.502026-08-06 · Output list $2.502026-08-07 · Output list $2.502026-08-08 · Output list $2.502026-08-09 · Output list $2.502026-08-10 · Output list $2.502026-08-11 · Output list $2.502026-08-12 · Output list $2.502026-08-13 · Output list $2.502026-08-05 · Min blended $0.8532026-08-06 · Min blended $0.8532026-08-07 · Min blended $0.8532026-08-08 · Min blended $0.8532026-08-09 · Min blended $0.8532026-08-10 · Min blended $0.8532026-08-11 · Min blended $0.8532026-08-12 · Min blended $0.8532026-08-13 · Min blended $0.853