DataLearner logo
KI

Kimi K2 0905

Retired · 2026-05-25Chat modelKimi K2Kimi K2

Kimi K2-Instruct-0905

Release date: 2025-09-05Updated: 2026-09-04Retired: 2026-05-25Views: 1,577
Parameters
1T
Context length
256K
Multilingual
Supported
Reasoning ability
4/5

Kimi K2-Instruct-0905 is an AI model published by Moonshot AI, released on 2025-09-05, for Chat model, with 1T parameters, and 256K context length, requiring about 1.01TB storage, with a 75.20 score on AIME2025.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

Kimi K2 0905

Model basics

Reasoning traces
No data
Thinking modes
Standard Mode (Default)Thinking Mode
Context length
256K tokens
Max output length
4K tokens
Model type
Chat model
Modality (in / out)
Text → Text
Release date
2025-09-05
Model file size
1.01TB
MoE architecture
Yes
Total params / Active params
1T / 32B
Knowledge cutoff
No data
Kimi K2 0905

Open source & experience

Code license
Weights license
MIT License- Commercial use permitted
GitHub repo
N/A
Live demo
Kimi K2 0905

Official resources

Paper
N/A
DataLearnerAI blog
Kimi K2 0905

API details

API speed
2/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Standard
TypeConditionInputOutput
Text-$0.600/ 1M tokens$2.50/ 1M tokens
Batch
TypeConditionInputOutput
Text-$0.300/ 1M tokens$1.25/ 1M tokens
Cache PricingPrompt Cache
TypeTTLWriteRead
Text-$0.600/ 1M tokens$0.300/ 1M tokens
Text-$0.600/ 1M tokens
Cache = write
Text-$0.300/ 1M tokens
Cache = hit

“—” means the modality is not billed in that direction, or the vendor has not published a price for it.

Kimi K2 0905

Benchmark Results

Kimi K2 0905 currently shows benchmark results led by Terminal-Bench (6 / 35, score 44.50), SWE-bench Verified (65 / 116, score 69.20), AIME2025 (68 / 107, score 75.20). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking
Tool usage

General Knowledge

1 evaluations
Benchmark / mode
Score
Rank/total
HLE
Thinking ModeTools
21.70
133 / 197

Coding and Software Engineer

3 evaluations
Benchmark / mode
Score
Rank/total
SWE-bench Verified
Standard Mode
69.20
65 / 116
SWE-bench Verified
Thinking ModeTools
69.20
65 / 116
27.67
61 / 62

Math and Reasoning

2 evaluations
Benchmark / mode
Score
Rank/total
AIME2025
Thinking ModeTools
75.20
68 / 107
IMO-ProofBench
Standard Mode
7.10
13 / 16

AI Agent - Tool Usage

2 evaluations
Benchmark / mode
Score
Rank/total
Terminal-Bench
Standard Mode
44.50
6 / 35
Terminal-Bench
Thinking ModeTools
44.50
6 / 35
Kimi K2 0905

Publisher

Kimi K2-Instruct-0905

Model Overview

Kimi K2-Instruct-0905 is an AI model published by Moonshot AI, released on 2025-09-05, for Chat model, with 1T parameters, and 256K context length, requiring about 1.01TB storage, with a 75.20 score on AIME2025.

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code