Claim · 2025-07-25 · publication · Stated in source · Supported

A StepFun team posted the Step-3 report to arXiv on 25 July 2025 (version 1), describing Step-3, a 321-billion-parameter vision-language model.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

StepFun / 阶跃星辰 → published → Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

StepFun / 阶跃星辰published ↓Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Date: 2025-07-25 · publication Records dated 2025 →

Scope and limitsThe report is the developer's own description; performance and benchmark claims in it are out of scope. Credit to StepFun rests on the team line of the report, not on a separate statement of ownership.

The evidence 02

primary sourceSupports the claim

Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding (arXiv abstract page, v1)

arXiv · 2025-07-25

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“StepFun: Bin Wang, Bojun Wang”

Location in source arXiv abstract page, author list

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-31

Source details and check notes

Authors: StepFun
Version: Dated document, retrieved on 2026-10-29 (SHA-256 b4595fd9b76a831dbb922ec1d0c5471efff5f22d2315956fd9e6221d655f8508) directly from the publisher. Version: 2507.19427v1. Abstract page of version 1 (submitted 2025-07-25); the page lists no later version at retrieval, not inspected.
Canonical address: Open the canonical source ↗
Accessed: 2026-10-29
Language: English (en)
Attribution: StepFun, “Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding (arXiv abstract page, v1)”, 2025-07-25.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, Chinese frontier-AI institutions, third collection (research/china-labs-3). The quotation (5 words) matched the text retrieved for the candidate pack on 2026-10-29 and again in a fresh retrieval the same day, read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. No human editorial review.

primary sourceSupports the claim

Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding (arXiv abstract page, v1)

arXiv · 2025-07-25

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“This paper introduces Step-3, a 321B-parameter VLM”

Location in source arXiv abstract page, abstract

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-31

Source details and check notes

Authors: StepFun
Version: Dated document, retrieved on 2026-10-29 (SHA-256 b4595fd9b76a831dbb922ec1d0c5471efff5f22d2315956fd9e6221d655f8508) directly from the publisher. Version: 2507.19427v1. Abstract page of version 1 (submitted 2025-07-25); the page lists no later version at retrieval, not inspected.
Canonical address: Open the canonical source ↗
Accessed: 2026-10-29
Language: English (en)
Attribution: StepFun, “Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding (arXiv abstract page, v1)”, 2025-07-25.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, Chinese frontier-AI institutions, third collection (research/china-labs-3). The quotation (7 words) matched the text retrieved for the candidate pack on 2026-10-29 and again in a fresh retrieval the same day, read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. Publication of a specific document or version. Publication of a policy is not evidence of implementation.

Date. Submission date of the cited arXiv version; not the model's release date and not the date of later versions.

Record ID c3-stepfun-step-3-report