Back to Explore
StepFun: Step 3.7 Flash
StepFun·Released 2026-05-28
256K ctxMultimodal
Comparison data ready93% coverage16/16 fields directly observedUpdated Jul 13, 2026, 10:31 PM
About
Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model. It pairs a 196B-parameter language backbone with a vision encoder for native image and video understanding, activating roughly 11B parameters...
Related Models
Pricing
Input
$0.20
per 1M tokens
Output
$1.15
per 1M tokens
Blended
$0.44
per 1M tokens
Cheaper than 57% of models. Median price is $0.67/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.44
Monthly
$13.13
vs. Similar Models
Google: Gemma 4 26B A4B Q:0.0
$0.13-71%
DeepSeek V3.1 Terminus (Reasoning)Q:+0.1
$1.91+337%
OpenAI: o3Q:+0.1
$3.50+700%
GPT-5.4 nano (medium)Q:-0.1
$0.46+6%
Performance
407
tokens/sec
Faster than 98% of models
0.67
seconds
Faster than 72% of models
5.59
seconds
Faster than 47% of models
Market Median
99 tok/s
311% faster
Median TTFT
1.11s
39% faster
Throughput/Dollar
929
tok/s per $/1M
Speed Comparison
Granite 3.3 8B (Non-reasoning)
404 tok/s-1%
Granite 4.0 H Small
426 tok/s+5%
LFM2.5-VL-1.6B
430 tok/s+6%
Context Window
256K
tokens
Larger than 56% of models
Max Output
256K
tokens
100% of context
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
80.9%
HLE
19.9%
LiveCodeBenchNot evaluated
SciCode
40.0%
TerminalBench Hard
35.6%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025Not evaluated
IFBench
67.3%
Long Context Recall
63.7%
Tau2
98.5%
Market AverageTop Score