Step 3.5 Flash

By price, Step 3.5 Flash ranks 6th cheapest of 68 on input and 12th on output. Its quality score of 79 ranks 53rd of 68, and on value (quality score ÷ output price) it ranks 12th of 68. Its blended rate sits 95% below the catalog average.

In real billing terms: a support chatbot handling 50,000 turns a month (2,000 input / 500 output tokens each) costs about $18 per month on Step 3.5 Flash, while a RAG workload of 20,000 queries (8,000 input / 800 output tokens each) runs about $21. Because output is priced 3.0× input here, long-generation tasks scale cost fastest — cap output length before you optimise anything else.

A 262K context window holds roughly 349 pages of A4 text (at ~750 tokens per page), ranking 33rd of 68 here. That covers typical RAG and multi-turn chat, though book-length inputs still need retrieval and chunking. Output runs at about 190 tok/s (9th fastest of 68), fast enough for real-time interaction and high-throughput batches.

We track 2 StepFun models, and Step 3.5 Flash's blended rate sits 49% below that provider's average. If your task tolerates a little less headroom, qwen3.5-flash (Alibaba) blends about 63% cheaper at a quality score of 82, and is worth benchmarking first. Nothing in the catalog scores higher on quality, so this is the ceiling.

The closest model in StepFun's line-up is Step 3.7 Flash ($0.20 input / $1.15 output, quality 85). Against it, Step 3.5 Flash scores 6 points lower on quality, runs 50 tok/s faster (about 36%), costs 192% less blended. When two models sit this close on price, the deciding factor is usually the quality ceiling — run both against your real prompts before picking.