25 Sept 2026
Will social media survive the AI change? What the evidence says
Our overnight sample had no social referrers, while Meta's crawlers made 5.9% of page requests. Usage, labelling rules and the EU AI Act…
A 600B mixture-of-experts flagship with 27B active per token and a million-token window, at $1.00 per million in and $2.70 out — priced well under the frontier it is measured against.
Step 5 Preview is StepFun's flagship model for agentic work, announced on 20 September 2026 with the API and the company's studio opening the same day. It is a sparse mixture-of-experts model with 600 billion total parameters and 27 billion active per token, a context window the vendor documents as 1M tokens, and native image and video input with text output. StepFun's own pricing table puts it at $1.00 per million input tokens on a cache miss, $0.05 per million on a cache hit — a 95% discount — and $2.70 per million output tokens, with the note that output tokens include the reasoning the model does before it answers, which is the charge buyers of reasoning models most often underestimate. The measured figure worth having is not the vendor's: Artificial Analysis scores it 44 on its Intelligence Index, placing it 24th of the 200 models it tracks, and reports 99.8 output tokens a second and 2.96 seconds to the first token on StepFun's own API. It is also a verbose model by that harness's account, generating 160M output tokens across the index, which matters when output is the expensive side of the bill. Two caveats belong with it. The open weights are a promise rather than a fact: StepFun's materials put a BF16 release on 15 October 2026, and the Hugging Face repository that carries the name holds a single .gitattributes file with no weights, no licence, no model card and no configuration. And the release date here is 20 September, the day StepFun announced it and opened the API; Artificial Analysis dates the model 18 September, which is when it measured it. Third-party configurations listing a 350,000-token window against the announced 1M have also been reported and are not resolved by the vendor's documentation, which states 1M.
| Benchmark | Official | Community avg |
|---|---|---|
| Artificial Analysis Intelligence Index v4.3.2llm | 44 index | — |
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
An earned signal from verification, reviews, awards, transparency and engagement — the vendor can't buy it.
Updated 9/20/2026
Used it? Your experience helps other buyers decide.
Write a reviewNo questions yet. Be the first to ask about Step 5 Preview.
Everything here links back to the same verified catalogue. Pick your next stop.