On 29 May 2026, StepFun announced Step 3.7 Flash. StepFun targeted high-frequency production agents with native vision, selectable reasoning levels and unusually high claimed throughput.
This release brief was checked against first-party material on 10 August 2026. The date above is the public announcement date, not the date a repository was created or a third-party provider added the model. Where access or weights arrived later, that distinction is recorded below.
Release record
| Field | Verified detail |
|---|---|
| Announcement | 29 May 2026 |
| Availability or weight release | 29 May 2026 by API, with open weights published on Hugging Face |
| Release type | open-weight multimodal agent model |
| Access | Apache 2.0 weights plus StepFun API and partner endpoints |
| Architecture | a 198B sparse vision-language MoE with about 11B active parameters and a 1.8B vision encoder |
| Maximum stated context | 256K tokens |
What changed
Step 3.7 Flash is one of the clearest efficiency challengers in the window. The official card ties approximately 11B active parameters to throughput up to 400 tokens per second, while publishing results across visual search, tool orchestration and repository work. The trade-off is visible too: its Terminal-Bench and GDPval results trail its strongest visual and orchestration scores, which helps teams decide where the model is most credible.
The practical comparison is therefore not simply whether Step 3.7 Flash has the largest headline score. Teams need to ask whether its architecture, access terms, latency, tool behaviour and evaluation setup match the workload they actually intend to run. A model can lead one harness while losing on cost, refusal behaviour, multilingual quality or repeatability in another.
Benchmarks worth retaining
| Evaluation | Reported result | How to read it |
|---|---|---|
| SimpleVQA with search | 79.2 | Visual grounding plus retrieval result |
| ClawEval 1.1 | 67.1 | Long-horizon agent execution score |
| SWE-Bench Pro | 56.3 | Repository-level software engineering result |
These are release-time results, not independently reproduced guarantees. Throughput depends heavily on hardware, batching and serving software, while several agent results use the lab’s preferred harness and reasoning setting. Scores should remain attached to the disclosed effort setting, agent harness, tool access, timeout, context-management policy and judge model. Moving a number into a procurement sheet without those conditions creates false comparability.
Architecture and access
Step 3.7 Flash is described as a 198B sparse vision-language MoE with about 11B active parameters and a 1.8B vision encoder with 256K tokens of stated context. Its access position at verification time is Apache 2.0 weights plus StepFun API and partner endpoints. That wording matters: open weights, source-available weights, an API, a product preview and a research demonstration give adopters very different rights and different levels of reproducibility.
Before deployment, record the exact model identifier or checkpoint, inference stack, quantisation, reasoning setting, region, price schedule and supplier terms. If the release uses a custom licence, read the licence itself rather than relying on the word “open” in launch copy. If it is API-only, preserve the dated documentation and change-notice route because the served snapshot can change without a downloadable artefact.
What an evaluation should test next
For Step 3.7 Flash, a credible internal gate should include:
- a frozen set of representative tasks with pass, fail and abstain criteria;
- a matched baseline using the same tools, timeout, prompt budget and reviewer rubric;
- repeated runs to expose variance rather than reporting a single best attempt;
- latency, token use and total task cost alongside task success;
- adversarial, multilingual and long-context cases relevant to the real deployment; and
- rollback evidence showing the previous model can be restored safely.
The wider model change-control guide explains how to keep model, prompt, tool and corpus changes reconstructable. The AI dependency inventory guide covers the release and supplier records needed after deployment.
AIEngine verdict
Step 3.7 Flash deserves a serious open-model trial for visual agents and high-volume tool workflows. Keep its speed claim hardware-specific and compare failure recovery, not only aggregate task success.
This is a launch assessment, not a certification. Benchmark leadership is useful evidence of where to test; it is not authorization to place the model in a high-impact workflow without domain evaluation, security review and an accountable owner.
Primary sources
Image provenance
Hero image: StepFun official model card via Hugging Face. The locally served WebP is a crop of the first-party release or model-card asset recorded in the repository provenance manifest.
