Saltar al contenido principal

Kimi K2 Thinking

KimiKimiOpen WeightMIT · Uso Comercial

Descripción

Kimi K2 Thinking is the latest, most capable version of open-source thinking model. Starting with Kimi K2, it is built as a thinking agent that reasons step-by-step while dynamically invoking tools. It sets a new state-of-the-art on Humanity's Last Exam (HLE), BrowseComp, and other benchmarks by dramatically scaling multi-step reasoning depth and maintaining stable tool-use across 200–300 sequential calls. At the same time, K2 Thinking is a native INT4 quantization model with 256k context window, achieving lossless reductions in inference latency and GPU memory usage. Key features include deep thinking & tool orchestration with end-to-end training to interleave chain-of-thought reasoning with function calls, native INT4 quantization via Quantization-Aware Training (QAT) achieving lossless 2x speed-up, and stable long-horizon agency maintaining coherent goal-directed behavior across up to 200–300 consecutive tool invocations.

Fecha de lanzamiento
2025-11-06
Parámetros
1.0T
Longitud del contexto
262K
Modalidades
text

Radar de capacidades

70
general
50
coding
80
reasoning
51
scienceest.
50
agents
0
multimodal

Science se estima a partir de las puntuaciones científicas de LLM Stats o del razonamiento cuando no hay benchmarks científicos dedicados.

Rankings

No hay datos de ranking disponibles

Puntuaciones de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

Finance

FinSearchComp-T347.4%Aut.

Healthcare

HealthBench58.0%Aut.

Language

MMLU-Redux94.4%Aut.
MMLU-Pro84.6%Aut.

Math

AIME 2025100.0%Aut.
HMMT 202597.5%Aut.
IMO-AnswerBench78.6%Aut.

Reasoning

FRAMES87.0%Aut.
GPQANYU + Cohere + Anthropic (2023)84.5%Aut.
LiveCodeBench v683.1%Aut.
SWE-Bench Verified71.3%Aut.
BrowseComp-zh62.3%Aut.
SWE-bench Multilingual61.1%Aut.
BrowseCompOpenAI (2025)60.2%Aut.
Seal-056.3%Aut.
Humanity's Last Exam51.0%Aut.
OJBench48.7%Aut.
Terminal-Bench47.1%Aut.
SciCode44.8%Aut.
Multi-SWE-Bench41.9%Aut.

Writing

WritingBench73.8%Aut.

Índices de evaluación AA

(Artificial Analysis)

No hay datos de evaluación AA disponibles

Puntuaciones por categoría LLM Stats

(LLM Stats (zeroeval))
Language
90
Legal
80
Math
80
Finance
80
Reasoning
70
Search
70
Frontend Development
70
General
70
Healthcare
70
Communication
70
Creativity
70
Writing
70
Physics
60
Biology
60
Chemistry
60
Agents
50
Code
50
Vision
50

Precios

Precio de entrada$0.6 / 1M tokens
Precio de salida$2.5 / 1M tokens
Precio mixto (3:1)$1.075 / 1M tokens
Precio de lectura caché$0.15 / 1M tokens

Velocidad

Tokens/seg122.3
Retraso del primer token1.53s
Tiempo hasta la respuesta17.87s

Ranking de Precios por Proveedor

Ranking de Precios por Proveedor

11 proveedores

Más barato: OpenCodeMás caro: AI/ML API
ProveedorEntradaSalida
1OpenCodeMás barato
$0.4
$2.5
2Helicone
$0.48
$2
3Alibaba (China)
$0.574
$2.294
4302.AI
$0.575
$2.3
5Qiniu
$0.57971
$2.31884
6KimiPRINCIPAL
$0.6
$2.5
7NanoGPT
$0.6
$2.5
8OpenRouter
$0.6
$2.5
9Novita AI
$0.6
$2.5
10Kilo Gateway
$0.6
$2.5
11AI/ML API
$0.6348
$2.645

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas