ModelPriceWatch.com
Last scan 2026-11-22 Models tracked 322 Providers 42 Cheapest paid pplx-decider-flash-v1.3 $0.010/Mtok in Every price links to its source

Step 3.5 Flash

by StepFun · 196B (A11B) parameters

Current budget open weights cheap tier

Today's price · per 1M tokens

Input

$0.100

per 1M tokens

Output

$0.300

per 1M tokens

Blended

$0.150

blended $/1M — 3:1 weighted input:output

Cached input

$0.020

20% of input — prompt caching

Source: official StepFun pricing · read MODELPRICEWATCH.COM · 2026-11-22

Price receipt

We read StepFun's own pricing page on and found step-3.5-flash listed with a price on the same row — that read is where the number above comes from. We keep the page text we read; its content hash is 74bc070d2d.

Overview

StepFun's text-only Flash reasoning model, a 196B-parameter sparse MoE with 11B active per token and a 256K-token context, built for fast tool calling and multi-step agent tasks. StepFun released the weights under Apache 2.0 and sells it on its own Open Platform, alongside an agent-tuned snapshot, step-3.5-flash-2603, at the same rate. $0.10/$0.30 per 1M tokens.

Run it yourself

Deploy this open model on rented GPUs

Open weights mean you can self-host instead of paying per-token API prices. These platforms let you serve it on demand — often cheaper at scale.

Partner links — we may earn a commission if you sign up, at no cost to you. We only list platforms we'd recommend regardless.

Capabilities

struck through = not supported
Input 1/4
Text ✓ Image Audio Video
Output
Text ✓
Features 3/9
Prompt caching ✓ Reasoning Coding Fast inference Long context ✓ Open weights ✓ Multimodal Web search Realtime

Benchmark performance

accuracy % · higher is better
Percentile vs all tracked models 19.2th1 independent measurement
Percentile per $/Mtok 128

We hold no per-benchmark accuracy scores for this model yet, so it has no accuracy average. That is a gap in our coverage, not a sign the model is untested — independent evaluators often publish a composite index for a new model long before releasing its per-benchmark numbers. The percentile above is its standing across the independent composites below.

Independent composite scores
  • Arena (arena.ai): 1392.8 ELO (Human preference)

Source: Arena (arena.ai) leaderboard dataset, CC BY 4.0

Artificial Analysis evaluations
  • Humanity's Last Exam: 21.1% (measured 2026-11-22)
  • GPQA Diamond: 83.1% (measured 2026-11-22)
  • AA-LCR (long-context reasoning): 50% (measured 2026-11-22)
  • CritPt (research-level physics): 2.5% (measured 2026-11-22)
  • IFBench (instruction following): 64.6% (measured 2026-11-22)
  • τ²-Bench Telecom: 94.4% (measured 2026-11-22)
  • Terminal-Bench Hard: 27.3% (measured 2026-11-22)

Source: Artificial Analysis (artificialanalysis.ai)

How it stacks up
  • Ranks #166 of 207 comparably-measured models by percentile score, across 1 independent measurement
  • Ranks #22 of 207 comparably-tested models by normalized performance per dollar

Source: · updated · See full rankings →

Specifications

Provider
StepFun
Context window
262K tokens
Modality
text
Parameters
196B (A11B)
Open source
Yes — open weights available
Released
Status
Current
Last updated
Tags
budgetopen-weightsmoecaching

Availability verified: — listed on StepFun's own page