K2-V2 (medium)
Intelligence
18.7
#284/532
Coding
14
#310/88
Math
64.7
#107/265
Speed
—
Pricing
—
K2-V2 (medium) is the MBZUAI Institute of Foundation Models’ offering designed for various applications in AI. It features an Intelligence Index of 18.7 and a Math Index of 64.7, making it suitable for tasks that require analytical reasoning and problem-solving.
When to Use K2-V2 (medium)
✓ Best For
- Data analysis and interpretation.
- Mathematical problem-solving.
- AI-driven coding assistance.
✗ Not Ideal For
- High-speed processing tasks due to unspecified speed.
- Applications requiring extensive coding capabilities.
How K2-V2 (medium) Compares
Intelligence Index · Higher is better
MistralNVIDIAMBZUAI Institute of Foundation ModelsTrillion LabsAnthropic
Benchmark Profile
Math Index
AlibabaInclusionAIMBZUAI Institute of Foundation ModelsDeepSeekNVIDIA
Intelligence · Coding · Math
Intelligence
Coding
Math
All Benchmark Scores (13)
| Benchmark | Score | |
|---|---|---|
| Intelligence Index | 18.7 | |
| Coding Index | 14 | |
| Math Index | 64.7 | |
| MMLU-Pro | 761% | |
| GPQA | 598% | |
| LiveCodeBench | 541% | |
| HLE | 44% | |
| SciCode | 25.2% | |
| IFBench | 55.1% | |
| LCR | 28% | |
| TerminalBench Hard | 8.3% | |
| Tau2 | 24.9% | |
| AIME 2025 | 64.7% |
Data: Artificial Analysis · Updated: March 27, 2026
Frequently Asked Questions (15)
When was K2-V2 (medium) released?
K2-V2 (medium) was released on December 5, 2025.
Who created K2-V2 (medium)?
K2-V2 (medium) was created by MBZUAI Institute of Foundation Models.
How intelligent is K2-V2 (medium)?
K2-V2 (medium) scores 19 on the Artificial Analysis Intelligence Index, placing it above average among other open weight models of similar size (median: 15).
How verbose is K2-V2 (medium)?
When evaluated on the Intelligence Index, K2-V2 (medium) generated 16M output tokens, which is somewhat higher than average compared to other open weight models of similar size (median: 7.3M).
Is K2-V2 (medium) a reasoning model?
Yes, K2-V2 (medium) is a reasoning model. It uses extended thinking or chain-of-thought reasoning to work through complex problems before providing an answer.
What input modalities does K2-V2 (medium) support?
K2-V2 (medium) supports text input.
What output modalities does K2-V2 (medium) support?
K2-V2 (medium) supports text output.
Can K2-V2 (medium) process images?
No, K2-V2 (medium) does not support image input. It can only process text.
Is K2-V2 (medium) multimodal?
No, K2-V2 (medium) is not multimodal. It only supports text input.
What is the context window of K2-V2 (medium)?
K2-V2 (medium) has a context window of 510k tokens. This determines how much text and conversation history the model can process in a single request.
Is K2-V2 (medium) open source?
Yes, K2-V2 (medium) is open weights. The model weights are publicly available and can be downloaded for self-hosting.
How many parameters does K2-V2 (medium) have?
K2-V2 (medium) has 70 billion parameters.
What is the license for K2-V2 (medium)?
K2-V2 (medium) is released under the Apache 2.0 license. This license allows commercial use.
How does K2-V2 (medium) perform on benchmarks?
K2-V2 (medium) achieves a score of 19 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.
Is K2-V2 (medium) available via API?
K2-V2 (medium) is an open weights model that can be self-hosted.
