MODEL DIRECTORY
Follow the model.
See the evidence.
Benchmark results, evaluation settings and disclosed training methods—collected in one place for each model.
3,335 reported models and checkpoint variants. Scores keep their original units and sources.
99 models · page 2 of 3
Kimi
; $5.426 per task; 70 median tokens per secondTogether AIKimi K3 / Fireworks
1 benchmarks · 1 observations
Explore results →Kimi
; $5.975 per task; 48 median tokens per secondFireworksKimi K3 / Moonshot
1 benchmarks · 1 observations
Explore results →Kimi
Kimi (Deep Think)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K2 Instruct (Terminus 2)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K2 Thinking (Terminus 2)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K2.7 Code (Terminus 2)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K3 (Moonshot)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K3 (reasoning on)
1 benchmarks · 2 observations
Explore results →Kimi
Kimi K3 (Terminus 2)
1 benchmarks · 1 observations
Explore results →Kimi
Kimi K3.1 Pro / Kimi Code
1 benchmarks · 2 observations
Explore results →Kimi
Kimi research checkpoint without cyber risk reduction / Kimi Code
1 benchmarks · 2 observations
Explore results →Kimi
Kimi-Audio-7B-Instruct*
1 benchmarks · 4 observations
Explore results →Kimi
kimi-k2-0905
1 benchmarks · 1 observations
Explore results →Kimi
Kimi-K2.5 (Mini-SWE)
1 benchmarks · 1 observations
Explore results →Kimi
kimi-k2p5
1 benchmarks · 1 observations
Explore results →Kimi
kimi-k3-max
1 benchmarks · 1 observations
Explore results →Kimi
kimi-k3-max-128k†
1 benchmarks · 1 observations
Explore results →Kimi
KimiChat-200k
1 benchmarks · 1 observations
Explore results →Microsoft
Microsoft-Decision-1
1 benchmarks · 1 observations
Explore results →Kimi
moonshotai/kimi-k2.5@reasoning=high
1 benchmarks · 4 observations
Explore results →Kimi
moonshotai/kimi-k2.5@reasoning=none
1 benchmarks · 4 observations
Explore results →Kimi
moonshotai/kimi-k2.6@reasoning=none
1 benchmarks · 2 observations
Explore results →Kimi
moonshotai/kimi-k2.6@reasoning=xhigh
1 benchmarks · 2 observations
Explore results →Kimi
moonshotai/kimi-k2@reasoning=default
1 benchmarks · 4 observations
Explore results →Kimi
moonshotai/kimi-k3@reasoning=minimal
1 benchmarks · 4 observations
Explore results →Kimi
moonshotai/kimi-k3@reasoning=xhigh
1 benchmarks · 4 observations
Explore results →Microsoft
phi-1.5 (1.3B)
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3 (7B)
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-medium-128k
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-Medium-4k-Instruct
1 benchmarks · 2 observations
Explore results →Microsoft
Phi-3-mini
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-small 7b*
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-small-128k-instruct
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-Vision
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3-Vision-Instruct
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3.5-4B
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3.5-mini-instruct
1 benchmarks · 6 observations
Explore results →Microsoft
Phi-3.5-mini-Instruct (Abdin et al., 2024)
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3.5-MoE-Instruct (Abdin et al., 2024)
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3.5V
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-3V
1 benchmarks · 1 observations
Explore results →Microsoft
Phi-4-multimodal-instruct*
1 benchmarks · 2 observations
Explore results →Microsoft
Phi3-14B-128k-Inst
1 benchmarks · 3 observations
Explore results →Microsoft
Phi3-7B-128k-Inst
1 benchmarks · 3 observations
Explore results →Microsoft
Phi3-V
1 benchmarks · 6 observations
Explore results →Microsoft
PubMedBERT Microsoft Research (Gu et al. 2020)
1 benchmarks · 1 observations
Explore results →StepFun
step-3.7-flash-medium
1 benchmarks · 1 observations
Explore results →StepFun
stepfun/step-3.5-flash@reasoning=minimal
1 benchmarks · 4 observations
Explore results →