MB
Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)
Language model · United Arab Emirates ·Sep 2025

K2 Think

Language Open (Restricted)
32B
Parameters
8
Benchmarks

About K2 Think

K2 Think is an AI model developed by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), in the language category, released in 2025, made available as an open-weights model with 32B params.

On this page you'll find K2 Think's full specifications, including 8 benchmark results. Review provider pricing and benchmark scores below, or compare K2 Think head-to-head with other language models.

→ More models by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) → All language models → AI benchmarks leaderboard

Links

Announcement ↗ Hugging Face ↗

Run K2 Think locally — what to buy

Share this build ↗

Sized to what K2 Think actually needs (~8k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Good
AMD Radeon RX 7900 XTX
Consumer GPU · ~$999
~41.3 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Better
NVIDIA GeForce RTX 4090
Consumer GPU · ~$1,599
~48.9 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Best
NVIDIA GeForce RTX 5090
Consumer GPU · ~$1,999
~87 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Consumer GPU
11–87 tok/s
9 options →
Apple
13.7–36.6 tok/s
17 options →
Datacenter GPU
37.1–120+ tok/s
13 options →

Community

Would you run K2 Think again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

K2 Think benchmark scores▶
Compare on the benchmark table →
Benchmark Category Score Variant
LiveCodeBench v5 code 63.97% avg@16
SciCode code 39.2% sub-problems
AIME 2024 math 90.83% avg@16
AIME 2025 math 81.24% avg@16
HMMT February 2025 math 73.75% avg@16
Omni-MATH-HARD math 60.73% avg@16
GPQA Diamond reasoning 71.08% avg@16
Humanity's Last Exam reasoning 9.95% avg@16

More models like K2 Think

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is K2 Think? ▶

K2 Think is an AI model developed by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), in the language category, released in 2025. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created K2 Think? ▶

K2 Think was developed by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and released in 2025.

Is K2 Think open source or proprietary? ▶

K2 Think ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does K2 Think cost? ▶

Pricing for K2 Think depends on the provider. See the providers table on this page for the latest API rates.

How does K2 Think perform on benchmarks? ▶

K2 Think is benchmarked across 8 evaluations on GenAIList, including LiveCodeBench v5 (63.97%). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about K2 Think?