Benchmarks of K2 v213
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Coding1
Documents1
Legacy · saturated or no longer reported4
No benchmark matches
—per task?
price not verifiedprice
—output speed?
512Kcontext
Rankingat high effort?
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGPT-6 Lunalow+46.4—$0.005Function calling only with none · Chat Completions+3
OpenMuse Spark Openxhigh+80.2—$0.84Reasoning off leaves output unchanged · Self-hosted / llama.cpp+1
BetterGranite 4.2 3BthinkingOpen+40.9—$0.006Context: 512K → 128K
BetterGemma 4 E4BthinkingOpen+35.9—$0.008Context: 512K → 128KBetterGPT-5.6 Lunalow+40.9—$0.01ModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
GPT-6 Lunalow+46.4—$0.005Function calling only with none · Chat Completions+3Granite 4.2 3BthinkingOpen+40.9—$0.006Context: 512K → 128KGemma 4 E4BthinkingOpen+35.9—$0.008Context: 512K → 128KGPT-5.6 Lunalow+40.9—$0.01
Mistral Small 4thinkingOpen+25.2—$0.015Context: 512K → 262KModelScoreCostPer taskWhat changes
Muse Spark OpenxhighOpen+80.2—$0.84Reasoning off leaves output unchanged · Self-hosted / llama.cpp+1
Kimi K3maxOpen+75.3—$2.00Sampling parameters are fixed+3
Step 5 PreviewthinkingOpen+72—$1.03web_search not available · Chat Completions+2
GLM 5.3maxOpen+69.9—$2.01Reasoning cannot be disabled+3MMotif 3Open+65.8—$0.50 / $2Context: 512K → 262KModelScoreCostPer taskWhat changes
IK2 Horizon 375B A23BOpen+52.6—$1 / $4Training behavior may differ · Self-hosted trainingIK2 Horizon MoVA 36B A4BthinkingOpen+44.6—$1 / $4IK2 Horizon 7BthinkingOpen+39—$0.30 / $1.20IK2 Horizon 3.7BthinkingOpen+27—$0.30 / $1.20IK2 Think V2Open+15.3——Context: 512K → 262KIK2 Horizon 0.9BthinkingOpen-17.2—$0.30 / $1.20Context: 512K → 131KAll Institute of Foundation Models models →Run it yourself
Fits onhardware needs not verified yet
LicenceApache-2.0free for commercial use
Benchmarks
all categoriesChat & Knowledge
Humanity's Last Exam10.5#184
AA Omniscience Index-55.6#239
CritPt (physics)0#180
AA Intelligence Index 4.3.2≈9.9#210
AA Omniscience · Accuracy18.5#175
GPQA Diamond68.1#196
AIME 2025 · AA archived78.3#49
Coding
LiveCodeBench · AA archived69.4#49
Agents & Tool Use
τ²-bench · AA27.8#157
Writing & Language
IFBench (instructions)60.1#85
FAQ
How much does K2 v2 cost?›
There is no per-unit API price: price not verified.
How good is K2 v2, and what is it best at?›
#183 of 334 in Chat & Knowledge, #126 of 239 in Coding, #155 of 242 in Agents & Tool Use, #85 of 245 in Writing & Language.
Is K2 v2 better than MiniMax M2?›
On AA Omniscience Index, K2 v2 scores -47.3 vs -46.6 — about the same.
What is the context window of K2 v2?›
512K tokens.
When was K2 v2 released? Is there a newer version?›
Released 2025-12-05. It is the current version in its line.
Is K2 v2 open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run K2 v2 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can K2 v2 read images?›
No, it accepts text only.
What is K2 v2?›
K2 v2 is an open-weights language model from Institute of Foundation Models, released Dec 2025.
What are the best alternatives to K2 v2?›
GPT-6 Luna (better), Muse Spark Open (open weights).