StepFun

★ 3/5 · Model Provider

Chinese foundation model lab offering the Step-3 API with reduced KV-cache inference costs optimized for long-context and complex reasoning tasks.

Pros

Optimized KV-cache efficiency reducing inference costs
Step-3 model designed for long-context and complex reasoning
Chinese foundation model alternative to global providers

Cons

Pricing details not publicly disclosed
Chinese-focused
limited international documentation
unclear performance benchmarks vs. competitors

Cost

Pay-per-token API (USD per 1M tokens on the international platform): step-3.7-flash $0.20 input ($0.04 cache hit), $1.15 output; step-3.5-flash $0.10 input ($0.02 cache hit), $0.30 output; stepaudio-2.5-chat $1.50 in, $3.50 out; TTS $0.85 per 10,000 characters; ASR $0.022/hour; China platform prices in CNY (step-3.7-flash ¥1.35 in, ¥8.10 out per 1M)

Verdict

Best for researchers and developers optimizing for long-context inference costs using Chinese LLMs and willing to work with specialized KV-cache architectures.

Visit StepFun