stepfun
StepFun: Step 3.7 Flash
stepfun/step-3.7-flash
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
stepfun/step-3.7-flash is cataloged as a text-output multimodal Mixture-of-Experts model that accepts text, image, and video input, provides a 256K-token context, and requires reasoning at selectable low, medium, or high effort. StepFun describes a sparse 198B-parameter design activating about 11B parameters per token for high-throughput agentic work. Reported evaluations emphasize visual verification, tool orchestration, and coding, but the evidence pack contains vendor and catalog data rather than independent validation.[1][2]
Strengths
- The catalog and StepFun documentation align on long-context multimodal input, text output, sparse activation, and three reasoning levels, a profile aimed at high-frequency production workloads.[1][2]
- StepFun reports strong results on visual question answering with search, multi-step tool orchestration, and software-engineering evaluations; the catalog also exposes tool calling, structured outputs, and reasoning controls.[1][2]
- An Apache-2.0 release and documented support for StepFun platforms, OpenRouter, NVIDIA NIM, vLLM, SGLang, Transformers, and llama.cpp provide several managed and self-hosted evaluation paths.[1][2]
Limitations
- The evidence pack contains only OpenRouter catalog data and StepFun documentation, so independent evaluation, safety behavior, knowledge cutoff, and real-world reliability remain unknown and should be validated on the intended workload.[1][2]
- OpenRouter lists video input and describes video understanding, while the supplied StepFun documentation chiefly presents native image understanding; video quality and supported formats need direct testing.[1][2]
- StepFun acknowledges that some terminal and professional-deliverable evaluations trail the cohort's peak, and local GGUF deployment is documented with about 120 GB minimum and 128 GB recommended memory; confirm task fit and hardware feasibility.[2]
Where to use it
- Evaluate for high-volume agentic pipelines that combine search, API or tool calls, and long-document analysis; workload tests should verify trajectory integrity and latency or cost before adoption.[1][2]
- Consider for coding assistants or concurrent coding agents that inspect multi-file repositories and draft patches, while benchmarking against your own repositories before relying on the vendor-reported results.[2]
- Try for text outputs grounded in charts, GUIs, documents, or catalog-listed video inputs, with a small pilot to confirm visual accuracy, especially for video.[1][2]
Sources
- [1] OpenRouter catalog · StepFun: Step 3.7 FlashSources checked: Nov 13, 2026
- [2] Model documentation · StepFun: Step 3.7 FlashSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model. It pairs a 196B-parameter language backbone with a vision encoder for native image and video understanding, activating roughly 11B parameters...
Specifications
- Model ID
- stepfun/step-3.7-flash
- Canonical ID
- stepfun/step-3.7-flash-20260528
- Creator
- stepfun
- Added to OpenRouter
- 28.05.2026
- Context length
- 262,144
- Modality
- text+image+video->text
- Input modalities
- text, image, video
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.200 / out $1.15
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | reasoning: mandatory · default medium · efforts: high, medium, low · parameters: include_reasoning, reasoning, reasoning_effort |
| Coding | CC — below the B band | — |
| Agents | No measurement | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: mandatory · default medium · efforts: high, medium, low · parameters: include_reasoning, reasoning, reasoning_effort |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | input: video |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 39.6 | CC — below the B band | 13.11.2026 |
| Design Arena | models · 3dwin rate 39.7% · rank 88 | 1,111Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · asciiartwin rate 44.3% · rank 47 | 1,143Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 41.8% · rank 73 | 1,170Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · datavizwin rate 43.3% · rank 70 | 1,173Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · gamedevwin rate 37.2% · rank 77 | 1,144Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · svgwin rate 36.4% · rank 73 | 1,075Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 41.4% · rank 74 | 1,168Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · websitewin rate 43.7% · rank 72 | 1,187Elo | BB — next 30% of the cohort | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Novitafp8 | 262,144 | $0.200 | $1.15 | — | — | 100% |
| StepFunfp8 | 256,000 | $0.200 | $1.15 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.