StepFun: Step 3.7 Flash

Community sentiment

Mixed (+0) — 0 positive, 0 negative, 2 neutral across 2 community samples from X and the web, as of 2026-09-04.

Community reports Step 3.7 Flash is live with low latency (≈0.4s on a 30k-token session), up to ~140 tok/s, and pricing listed at $0.20/M in and $1.00/M out. Independent benchmarks note behavior/gameplay is broadly similar to other 3.7 Flash models, with pricing and iteration cited as the main differentiators.

• “Step 3.7 Flash is live on Tiyuvta. - warm follow-up on a 30k-token session: 0.4s - up to 140 tok/s, measured on the live endpoint - $0.20/M in, $1.00/M out OpenAI + Anthropic formats, tools included. "model": "stepfun/step-3.7-flash" https://t.co/c224XxRep8” — X
• “I ran a Flappybench test on the DSV4 Flash Fast with Qwen 3.7 Flash, Step 3.7 Flash I used /design command from Command Code. Overall, all the gameplay feels the same, but pricing and iteration make the difference. Reviewed gameplay, iteration and cost. → DSV4 Flash Fast: https:/” — X

RuntimeWire coverage

• Claude Opus 5 (Fast) tops Newsroom Reliability v0.2 benchmark
• Z.ai GLM 5.3 Flash tops Editorial Craft benchmark at 0.94
• Open-weight models surge past closed rivals in Vercel token traffic
• The Week Open Weights Went Multimodal