Timeline of StepFun

New model: step-3-vl-10b (StepFun)

step-3-5-flash scores 32.58% on Terminal-Bench

step-3-5-flash scores 87.43% on τ²-bench

step-3-5-flash scores 66.53% on IFBench

step-3-5-flash scores 24.47% on Humanity's Last Exam

step-3-5-flash scores 82.63% on GPQA

step-3-5-flash scores 16.96 on Artificial Analysis Intelligence Index

Step 3.7 Flash scores 35.61% on Terminal-Bench

Step 3.7 Flash scores 39.33% on Terminal-Bench

Step 3.7 Flash scores 98.54% on τ²-bench

Step 3.7 Flash scores 75.32% on MMMU-Pro

Step 3.7 Flash scores 67.28% on IFBench

Step 3.7 Flash scores 43.87% on SciCode

Step 3.7 Flash scores 21.41% on Humanity's Last Exam

Step 3.7 Flash scores 80.91% on GPQA

Step 3.7 Flash scores 19.48 on Artificial Analysis Intelligence Index

step-3-vl-10b scores 5.3% on Terminal-Bench

step-3-vl-10b scores 16.08% on τ²-bench

step-3-vl-10b scores 63.99% on MMMU-Pro

step-3-vl-10b scores 50.2% on IFBench

step-3-vl-10b scores 10.84% on Humanity's Last Exam

step-3-vl-10b scores 68.99% on GPQA

step-3-vl-10b scores 7.69 on Artificial Analysis Intelligence Index

Step 3.5 Flash scores 27.27% on Terminal-Bench

Step 3.5 Flash scores 94.44% on τ²-bench

Step 3.5 Flash scores 64.56% on IFBench

Step 3.5 Flash scores 21.13% on Humanity's Last Exam

Step 3.5 Flash scores 83.13% on GPQA

Step 3.5 Flash scores 16.62 on Artificial Analysis Intelligence Index

Step 3.5 Flash: context length changed from 262144 to 256000

New company: StepFun

OpenRouter lists Step 3.5 Flash at $0.1 in / $0.3 out per 1M tokens

OpenRouter lists Step 3.7 Flash at $0.2 in / $1.15 out per 1M tokens