Step-5-Preview vs Gemini 3.8 Flash

API pricing, benchmark indices and specs — updated 19 October 2026.

stepfun-ai · A · Open weights
Params
604.3 B
Context
—
30d
2 K
google · B · API-only
Input / 1M
$0.83
Output / 1M
$4.13
AA index
40.9
proprietary · 1M context

Specification comparison

Differences are highlighted; identical values are muted.

Specification Step-5-Preview Gemini 3.8 Flash
Deployment Open weights · also via API API-only (closed weights)
Parameters 604.3 B n/a — API-only
Architecture Step5ForConditionalGeneration n/a — API-only
Context length 1,000,000 1,048,576
Licence other Proprietary
Commercial use Unknown Per API terms
Languages — n/a — API-only
Downloads 30d 2,467 n/a — API-only
Downloads all time 2 K n/a — API-only
Quantizations on the Hub SAFETENSORS n/a — API-only
Gated No n/a — API-only
First seen on the Hub 2026-09-20 n/a — API-only
API input price / 1M tokens $1.00 $0.83
API output price / 1M tokens $2.70 $4.13
Est. cost per 1M input + 1M output tokens $3.70 $4.95
Cached input / 1M tokens $0.10 $0.083
AA Intelligence Index 43.7 40.9
AA Coding Index — 76.3
AA Agentic Index — 40.2
Max output tokens 64,000 65,536
API providers 4 2
Tool calling Yes Yes
Reasoning mode Yes Yes

API cost

Estimated cost of 1M input + 1M output tokens via OpenRouter

Step-5-Preview $3.70
Gemini 3.8 Flash $4.95

Step-5-Preview is 1.3× cheaper at list price.

VRAM side by side

On RTX 4090 · 24 GB · 8K context unless noted

Step-5-Preview · fp16 806.5 GB / 24 GB
❌ Won’t fit

Adjacent comparisons