Skip to content

Kimi · Model profile

Kimi K2 Thinking

Intelligence, Performance & Price Analysis

Source: Canonical slug: kimi-k2-thinking · Canonical model registry at build time

Routing verdict

Frontier research / complex reasoning

Reserve this model for the hardest requests; simpler or repetitive work should stay on cheaper routes.

Kimi K2 Thinking summary metrics

Intelligence
33 scoreKimi K2 Thinking scores 33 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other open weight models of similar size (median: 25).
Speed
160.2 output tokens/secKimi K2 Thinking generates output at 160.2 tokens per second (based on the median across providers serving the model), which is well above average compared to other open weight models of similar size (median: 61.7 t/s).
Latency
1.26s TTFTKimi K2 Thinking has a time to first token (TTFT) of 1.26s (based on the median across providers serving the model), which is very competitive compared to other open weight models of similar size (median: 2.01s).
Input Price
$0.60 / 1M tokensKimi K2 Thinking costs $0.60 per 1M input tokens (somewhat higher than average, median: $0.59) and $2.50 per 1M output tokens (somewhat higher than average, median: $2.20), based on the median across providers serving the model.
Output Price
$2.50 / 1M tokensKimi K2 Thinking costs $0.60 per 1M input tokens (somewhat higher than average, median: $0.59) and $2.50 per 1M output tokens (somewhat higher than average, median: $2.20), based on the median across providers serving the model.

Assessment

Routing verdict and task fit

Kimi K2 Thinking scores 33 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other open weight models of similar size (median: 25). Kimi K2 Thinking generates output at 160.2 tokens per second (based on the median across providers serving the model), which is well above average compared to other open weight models of similar size (median: 61.7 t/s). Kimi K2 Thinking costs $0.60 per 1M input tokens (somewhat higher than average, median: $0.59) and $2.50 per 1M output tokens (somewhat higher than average, median: $2.20), based on the median across providers serving the model.
Task fit assessment
WorkloadRatingNotes
Complex reasoning & agentic workflowsOptimalIntelligence score 33 places this model among top performers. Suitable for multi-step analysis and agentic loops.
High-volume chat & customer-facingOptimalOutput speed 160.2 tokens/sec and capable intelligence make this suitable for real-time chat at scale.
Latency-sensitive applicationsOptimalTTFT 1.26s — adequate latency for most interactive use cases.
Cost-sensitive pipelinesViableOutput pricing at $2.50 is premium. Route high-volume simple tasks to cheaper alternatives.

Evidence

Evidence charts

16 published comparison charts for this profile. Kimi K2 Thinking is marked with a diamond and bold label; each chart has a table view and lists its source and retrieval date.

AA-Omniscience Index

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: -20.5, row 16 of 20 in the published order.
  1. Claude Fable 5 (with fallback)40.1
  2. Gemini 3.5 Flash22.7
  3. GPT-5.5 (xhigh)20.1
  4. Grok 4.3 (high)18.3
  5. Kimi K2.66.4
  6. Muse Spark4.1
  7. GLM-5.2 (max)4
  8. MiMo-V2.5-Pro3.6
  9. GLM-5.11.9
  10. MiniMax-M31.4
  11. MiniMax-M2.70.683
  12. Nemotron 3 Ultra-0.833
  13. Command A+-4
  14. DeepSeek V4 Pro (max)-10
  15. Kimi K2.7 Code-10.7
  16. Kimi K2 Thinking (this model)-20.5
  17. DeepSeek V4 Flash (max)-22.9
  18. Nex-N2-Pro-27.6
  19. Qwen3.5 397B A17B-29.8
  20. DeepSeek R1 (Jan)-31.3

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Artificial Analysis Intelligence Index by Open Weights / Proprietary

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 32.7, row 19 of 20 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. MiMo-V2.5-Pro42.2
  10. Kimi K2.7 Code41.9
  11. Nex-N2-Pro41
  12. DeepSeek V4 Flash (max)40.3
  13. GLM-5.140.2
  14. MiniMax-M2.738.1
  15. Nemotron 3 Ultra37.8
  16. Grok 4.3 (high)37.6
  17. GLM-4.733.7
  18. Qwen3.5 397B A17B33.7
  19. Kimi K2 Thinking (this model)32.7
  20. Ring-2.6-1T30.6

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 32.7, row 19 of 20 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. MiMo-V2.5-Pro42.2
  10. Kimi K2.7 Code41.9
  11. Nex-N2-Pro41
  12. DeepSeek V4 Flash (max)40.3
  13. GLM-5.140.2
  14. MiniMax-M2.738.1
  15. Nemotron 3 Ultra37.8
  16. Grok 4.3 (high)37.6
  17. GLM-4.733.7
  18. Qwen3.5 397B A17B33.7
  19. Kimi K2 Thinking (this model)32.7
  20. Ring-2.6-1T30.6

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open) · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 27.8, row 19 of 20 in the published order.
  1. Nemotron 3 Ultra83.3
  2. MiMo-V2-Flash52.8
  3. DeepSeek V4 Pro (max)50
  4. DeepSeek V4 Flash (max)50
  5. GLM-5.2 (max)44.4
  6. GLM-5.144.4
  7. GLM-4.744.4
  8. GLM-4.644.4
  9. Qwen3 235B A22B 250744.4
  10. gpt-oss-120b (high)38.9
  11. MiMo-V2.5-Pro38.9
  12. Nex-N2-Pro38.9
  13. Qwen3.5 397B A17B38.9
  14. Ring-2.6-1T38.9
  15. Step 3.7 Flash38.9
  16. Command A+38.9
  17. Kimi K2.633.3
  18. MiniMax-M333.3
  19. Kimi K2 Thinking (this model)27.8
  20. Kimi K2.7 Code27.8

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Intelligence

Artificial Analysis Intelligence Index · Higher is better · Evaluation results measured independently by Artificial Analysis

12 models in the published order. Kimi K2 Thinking: 32.7, row 11 of 12 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. Kimi K2 Thinking (this model)32.7
  12. gpt-oss-120b (high)23.8

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Output Speed

Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 160 t/s, row 8 of 20 in the published order.
  1. Step 3.7 Flash410 t/s
  2. gpt-oss-120b (high)314 t/s
  3. Nemotron 3 Ultra250 t/s
  4. GLM-5.2 (max)218 t/s
  5. Command A+196 t/s
  6. Gemini 3.5 Flash192 t/s
  7. Grok 4.3 (high)164 t/s
  8. Kimi K2 Thinking (this model)160 t/s
  9. Ring-2.6-1T136 t/s
  10. GLM-4.7126 t/s
  11. DeepSeek V4 Flash (max)112 t/s
  12. MiniMax-M399 t/s
  13. Nex-N2-Pro92 t/s
  14. GPT-5.5 (xhigh)88 t/s
  15. GLM-5.180 t/s
  16. Kimi K2.676 t/s
  17. DeepSeek V4 Pro (max)72 t/s
  18. Claude Fable 5 (with fallback)71 t/s
  19. Qwen3 235B A22B 250761 t/s
  20. MiMo-V2.5-Pro53 t/s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Output speed context
Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.

Speed

Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis

11 models in the published order. Kimi K2 Thinking: 160 t/s, row 6 of 11 in the published order.
  1. gpt-oss-120b (high)314 t/s
  2. Nemotron 3 Ultra250 t/s
  3. GLM-5.2 (max)218 t/s
  4. Gemini 3.5 Flash192 t/s
  5. Grok 4.3 (high)164 t/s
  6. Kimi K2 Thinking (this model)160 t/s
  7. MiniMax-M399 t/s
  8. GPT-5.5 (xhigh)88 t/s
  9. Kimi K2.676 t/s
  10. DeepSeek V4 Pro (max)72 t/s
  11. Claude Fable 5 (with fallback)71 t/s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Output speed context
Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.

End-to-End Response Time

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 3.12 s, row 6 of 20 in the published order.
  1. Step 3.7 Flash1.22 s
  2. gpt-oss-120b (high)1.59 s
  3. Nemotron 3 Ultra2 s
  4. GLM-5.2 (max)2.3 s
  5. Command A+2.55 s
  6. Kimi K2 Thinking (this model)3.12 s
  7. GLM-4.73.97 s
  8. Ring-2.6-1T3.67 s
  9. Grok 4.3 (high)3.05 s
  10. MiniMax-M35.05 s
  11. Nex-N2-Pro5.42 s
  12. Gemini 3.5 Flash2.6 s
  13. Qwen3 235B A22B 25078.14 s
  14. MiMo-V2.5-Pro9.52 s
  15. GLM-4.610.18 s
  16. GLM-5.16.28 s
  17. DeepSeek V4 Flash (max)4.46 s
  18. Kimi K2.7 Code9.72 s
  19. MiniMax-M2.79.79 s
  20. Kimi K2.66.56 s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 12.48 s, row 6 of 20 in the published order.
  1. Step 3.7 Flash4.88 s
  2. gpt-oss-120b (high)6.37 s
  3. Nemotron 3 Ultra9.09 s
  4. Command A+10.18 s
  5. GLM-5.2 (max)9.19 s
  6. Kimi K2 Thinking (this model)12.48 s
  7. GLM-4.715.87 s
  8. Ring-2.6-1T14.68 s
  9. Grok 4.3 (high)0 s
  10. MiniMax-M320.19 s
  11. Nex-N2-Pro21.66 s
  12. Gemini 3.5 Flash0 s
  13. Qwen3 235B A22B 250732.55 s
  14. MiMo-V2.5-Pro38.08 s
  15. GLM-4.640.7 s
  16. Kimi K2.7 Code43.34 s
  17. GLM-5.147.62 s
  18. MiniMax-M2.748.19 s
  19. DeepSeek V4 Flash (max)50.09 s
  20. Kimi K2.658.43 s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Latency considerations
Latency includes thinking / reasoning time when applicable. Lower is better for interactive use cases.

Context Window

Context window: tokens limit · Higher is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking: 256k, row 16 of 20 in the published order.
  1. Gemini 3.5 Flash1M
  2. Claude Fable 5 (with fallback)1M
  3. GLM-5.2 (max)1M
  4. DeepSeek V4 Pro (max)1M
  5. Grok 4.3 (high)1M
  6. MiniMax-M31M
  7. MiMo-V2.5-Pro1M
  8. DeepSeek V4 Flash (max)1M
  9. GPT-5.5 (xhigh)922k
  10. Nemotron 3 Ultra262.1k
  11. Muse Spark262.1k
  12. Qwen3.5 397B A17B262.1k
  13. Ring-2.6-1T262.1k
  14. Step 3.7 Flash262.1k
  15. Nex-N2-Pro262k
  16. Kimi K2 Thinking (this model)256k
  17. Kimi K2.6256k
  18. Kimi K2.7 Code256k
  19. MiMo-V2-Flash256k
  20. Qwen3 235B A22B 2507256k

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Context window sizing
Larger context windows enable complex multi-turn tasks but may increase latency and cost for long inputs.

Cost per Intelligence Index Task

Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better · Evaluation results measured independently by Artificial Analysis

19 models in the published order. Kimi K2 Thinking is not marked in this published set of 19 models.
  1. MiMo-V2.5-Pro0.00466
  2. DeepSeek V4 Pro (max)0.0063
  3. gpt-oss-120b (high)0.00454
  4. MiniMax-M30.0139
  5. Grok 4.3 (high)0.0101
  6. Kimi K2.7 Code0.0203
  7. GLM-5.10.0426
  8. Nemotron 3 Ultra0.0209
  9. DeepSeek R1 (Jan)0.0153
  10. GLM-4.60.0144
  11. GLM-4.70.0157
  12. Qwen3.5 397B A17B0.022
  13. Ring-2.6-1T0.0239
  14. Kimi K2.60.0216
  15. MiMo-V2-Flash0.00812
  16. GLM-5.2 (max)0.0268
  17. Gemini 3.5 Flash0.0884
  18. GPT-5.5 (xhigh)0.107
  19. Claude Fable 5 (with fallback)0.367

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost per Task

Weighted average cost (USD) per Intelligence Index task · Lower is better · Evaluation results measured independently by Artificial Analysis

10 models in the published order. Kimi K2 Thinking is not marked in this published set of 10 models.
  1. DeepSeek V4 Pro (max)$0.045
  2. gpt-oss-120b (high)$0.061
  3. MiniMax-M3$0.125
  4. Grok 4.3 (high)$0.139
  5. Nemotron 3 Ultra$0.244
  6. Kimi K2.6$0.346
  7. GLM-5.2 (max)$0.370
  8. Gemini 3.5 Flash$0.586
  9. GPT-5.5 (xhigh)$0.863
  10. Claude Fable 5 (with fallback)$2.75

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis

19 models in the published order. Kimi K2 Thinking is not marked in this published set of 19 models.
  1. gpt-oss-120b (high)3.5
  2. MiMo-V2.5-Pro6.7
  3. DeepSeek V4 Pro (max)7.2
  4. MiniMax-M314.5
  5. Grok 4.3 (high)12.5
  6. GLM-4.615.6
  7. MiMo-V2-Flash11.4
  8. DeepSeek R1 (Jan)22
  9. Nemotron 3 Ultra33.5
  10. Ring-2.6-1T23.5
  11. Kimi K2.7 Code24.3
  12. Qwen3.5 397B A17B24.9
  13. GLM-4.716.6
  14. GLM-5.140.4
  15. GLM-5.2 (max)32.9
  16. Kimi K2.625
  17. Gemini 3.5 Flash114
  18. GPT-5.5 (xhigh)131.8
  19. Claude Fable 5 (with fallback)507.6

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens) · Evaluation results measured independently by Artificial Analysis

20 models; segments: Input, Output, Cache Hit. Kimi K2 Thinking: $3.70, row 13 of 20 in the published order.
  • Input
  • Output
  • Cache Hit
  1. MiMo-V2-Flash$0.400
  2. DeepSeek V4 Flash (max)$0.423
  3. gpt-oss-120b (high)$0.900
  4. DeepSeek V4 Pro (max)$1.31
  5. MiMo-V2.5-Pro$1.31
  6. Step 3.7 Flash$1.39
  7. MiniMax-M3$1.56
  8. MiniMax-M2.7$1.56
  9. GLM-4.6$2.75
  10. Ring-2.6-1T$2.80
  11. GLM-4.7$3.33
  12. Nex-N2-Pro$3.25
  13. Kimi K2 Thinking (this model)$3.70
  14. Nemotron 3 Ultra$3.60
  15. Grok 4.3 (high)$3.95
  16. Qwen3.5 397B A17B$4.20
  17. Kimi K2.6$5.11
  18. Kimi K2.7 Code$5.11
  19. GLM-5.2 (max)$6.06
  20. GLM-5.1$6.06

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Time per Intelligence Index Task

Weighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Kimi K2 Thinking is not marked in this published set of 20 models.
  1. Nemotron 3 Ultra1.4
  2. Grok 4.3 (high)1.6
  3. Command A+1.7
  4. Step 3.7 Flash1.9
  5. gpt-oss-120b (high)1.9
  6. Gemini 3.5 Flash2.2
  7. GLM-4.72.5
  8. GLM-5.2 (max)2.8
  9. Ring-2.6-1T2.9
  10. GPT-5.5 (xhigh)3
  11. MiniMax-M33.9
  12. Qwen3 235B A22B 25074.4
  13. Claude Fable 5 (with fallback)4.8
  14. Nex-N2-Pro4.8
  15. GLM-5.14.9
  16. Kimi K2.7 Code5
  17. Qwen3.5 397B A17B5.1
  18. MiniMax-M2.75.5
  19. GLM-4.65.9
  20. MiMo-V2.5-Pro6.1

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference · Evaluation results measured independently by Artificial Analysis

20 models; segments: Passive Params, Active Params. Kimi K2 Thinking: 1,000B, row 3 of 20 in the published order.
  • Passive Params
  • Active Params
  1. DeepSeek V4 Pro (max)1,600B
  2. MiMo-V2.5-Pro1,023B
  3. Kimi K2 Thinking (this model)1,000B
  4. Kimi K2.61,000B
  5. Kimi K2.7 Code1,000B
  6. Ring-2.6-1T1,000B
  7. GLM-5.2 (max)753B
  8. GLM-5.1744B
  9. DeepSeek R1 (Jan)685B
  10. Nemotron 3 Ultra550B
  11. MiniMax-M3428B
  12. Nex-N2-Pro397B
  13. Qwen3.5 397B A17B397B
  14. GLM-4.7357B
  15. GLM-4.6357B
  16. MiMo-V2-Flash309B
  17. DeepSeek V4 Flash (max)284B
  18. Qwen3 235B A22B 2507235B
  19. MiniMax-M2.7230B
  20. Command A+218B

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Architecture scale
Larger models generally perform better but cost more to serve. Active parameter count matters for inference efficiency.

Specifications

Architecture and limits

Kimi K2 Thinking technical specifications
SpecificationValueSource
Model typeOpen weightsfaq
ReasoningYesfaq
Input modalitiesKimi K2 Thinking supports text input.faq
Output modalitiesKimi K2 Thinking supports text output.faq
Context window260k tokensfaq
Open weights / sourceYes, Kimi K2 Thinking is open weights. The model weights are publicly available and can be downloaded for self-hosting.faq
ParametersKimi K2 Thinking has 1 trillion parameters (32 billion active).faq
Active parametersKimi K2 Thinking is a Mixture of Experts (MoE) model with 1 trillion total parameters, but only 32 billion active parameters are used during inference.faq
LicenseKimi K2 Thinking is released under the Modified MIT License license. This license allows commercial use.faq
API availabilityYes, Kimi K2 Thinking is available via API through 6 providers.faq

Method

Methodology & Provenance

This page is rendered from the normalized profile and page JSON for Kimi K2 Thinking.

Benchmark values are preserved as normalized; only layout, disclosure ordering, and typography are adjusted for readability.

Questions

Model FAQ and technical disclosures

Kimi K2 Thinking was released on November 6, 2025.

Kimi K2 Thinking was created by Kimi.

Kimi K2 Thinking scores 33 (estimated) on the Artificial Analysis Intelligence Index, placing it above average among other open weight models of similar size (median: 25).

Kimi K2 Thinking generates output at 160.2 tokens per second (based on the median across providers serving the model), which is well above average compared to other open weight models of similar size (median: 61.7 t/s).

Kimi K2 Thinking has a time to first token (TTFT) of 1.26s (based on the median across providers serving the model), which is very competitive compared to other open weight models of similar size (median: 2.01s).

Kimi K2 Thinking costs $0.60 per 1M input tokens (somewhat higher than average, median: $0.59) and $2.50 per 1M output tokens (somewhat higher than average, median: $2.20), based on the median across providers serving the model.

Kimi K2 Thinking costs $0.60 per 1M input tokens and $2.50 per 1M output tokens (based on the median across providers serving the model). For a blended rate (7:2:1 cache hit/input/output ratio), this is $0.79 per 1M tokens. Pricing may vary by provider.

Yes, Kimi K2 Thinking is a reasoning model. It uses extended thinking or chain-of-thought reasoning to work through complex problems before providing an answer.

Kimi K2 Thinking supports text input.

Kimi K2 Thinking supports text output.

No, Kimi K2 Thinking does not support image input. It can only process text.

No, Kimi K2 Thinking is not multimodal. It only supports text input.

Kimi K2 Thinking has a context window of 260k tokens. This determines how much text and conversation history the model can process in a single request.

Yes, Kimi K2 Thinking is open weights. The model weights are publicly available and can be downloaded for self-hosting.

Kimi K2 Thinking has 1 trillion parameters (32 billion active).

Kimi K2 Thinking is a Mixture of Experts (MoE) model with 1 trillion total parameters, but only 32 billion active parameters are used during inference.

Kimi K2 Thinking is released under the Modified MIT License license. This license allows commercial use.

Kimi K2 Thinking achieves a score of 33 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.

Yes, Kimi K2 Thinking is available via API through 6 providers.

Kimi K2 Thinking is available through 6 API providers.

Models

  • 01KimiOther Kimi models in the catalog.
  • 02LeaderboardsCertified metric ranks, when published.