StepFun model product
Step 3.5 Flash
5 Flash is a fast, cost-effective text model from StepFun, optimized for quick inference and efficient token usage.
Updated Sep 8, 2026. Default version: Step-3.5-Flash
LLMBoard Score59.4Step-3.5-Flash
Coverage40%7 benchmark families
Context window65.5KTokens
Official input price$0.10StepFun (Global) API
Step 3.5 Flash Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Step 3.5 Flash Benchmark Results
Benchmark scores for Step-3.5-Flash.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkTau-bench | Score88.20% | Rank01 | Participants6 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkIMO-AnswerBench | Score85.40% | Rank09 | Participants20 | Percentile57.89% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench v6 | Score86.40% | Rank12 | Participants62 | Percentile81.97% | EvidenceC | Evaluated |
| BenchmarkAIME 2025 | Score97.30% | Rank14 | Participants119 | Percentile88.98% | EvidenceC | Evaluated |
| BenchmarkBrowseComp | Score69.00% | Rank32 | Participants63 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkTerminal-Bench 2.0 | Score51.00% | Rank35 | Participants51 | Percentile32.00% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Verified | Score74.40% | Rank41 | Participants113 | Percentile64.29% | EvidenceC | Evaluated |
| BenchmarkLM Arena Text Factuality | Score1,417.15 rating | Rank93 | Participants121 | Percentile23.33% | EvidenceA | Evaluated |
| BenchmarkLM Arena Text | Score1,403.77 rating | Rank100 | Participants210 | Percentile52.63% | EvidenceA | Evaluated |
| BenchmarkLM Arena Text Style Control | Score1,394.40 rating | Rank115 | Participants210 | Percentile45.45% | EvidenceA | Evaluated |
Step 3.5 Flash Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenatext factuality | Categoryjapanese | Rank32 | Rating / score1,371.19 | Votes415 | ObservationsN/A | Result date |
| Arenatext factuality | Categorygerman | Rank33 | Rating / score1,390.44 | Votes603 | ObservationsN/A | Result date |
| Arenatext factuality | Categorykorean | Rank40 | Rating / score1,362.17 | Votes595 | ObservationsN/A | Result date |
| Arenatext factuality | Categorypolish | Rank42 | Rating / score1,392.86 | Votes850 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryfrench | Rank51 | Rating / score1,450.13 | Votes1,520 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryspanish | Rank54 | Rating / score1,419.45 | Votes1,383 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry mathematical | Rank60 | Rating / score1,429.45 | Votes2,230 | ObservationsN/A | Result date |
| Arenatext factuality | Categorychinese | Rank62 | Rating / score1,467.23 | Votes2,334 | ObservationsN/A | Result date |
| Arenatext factuality | Categorymath | Rank62 | Rating / score1,421.97 | Votes2,393 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryexpert | Rank69 | Rating / score1,446.19 | Votes3,722 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry legal and government | Rank69 | Rating / score1,432.29 | Votes3,307 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry medicine and healthcare | Rank71 | Rating / score1,447.42 | Votes2,946 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryrussian | Rank77 | Rating / score1,413.44 | Votes5,023 | ObservationsN/A | Result date |
| Arenatext | Categoryspanish | Rank79 | Rating / score1,418.97 | Votes1,986 | ObservationsN/A | Result date |
| Arenatext factuality | Categorycoding | Rank79 | Rating / score1,473.28 | Votes12,261 | ObservationsN/A | Result date |
| Arenatext | Categoryjapanese | Rank81 | Rating / score1,351.46 | Votes651 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryhard prompts english | Rank81 | Rating / score1,453.13 | Votes13,433 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry software and it services | Rank82 | Rating / score1,460.93 | Votes20,272 | ObservationsN/A | Result date |
| Arenatext | Categorygerman | Rank84 | Rating / score1,402.35 | Votes1,055 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryinstruction following | Rank84 | Rating / score1,410.73 | Votes16,155 | ObservationsN/A | Result date |
| Arenatext | Categorycoding | Rank85 | Rating / score1,436.66 | Votes14,324 | ObservationsN/A | Result date |
| Arenatext | Categoryexpert | Rank85 | Rating / score1,422.44 | Votes4,959 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry life and physical and social science | Rank85 | Rating / score1,440.50 | Votes7,253 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryenglish | Rank86 | Rating / score1,433.10 | Votes23,577 | ObservationsN/A | Result date |
| Arenatext | Categoryindustry mathematical | Rank87 | Rating / score1,419.32 | Votes2,927 | ObservationsN/A | Result date |
| Arenatext | Categoryindustry software and it services | Rank87 | Rating / score1,435.18 | Votes21,517 | ObservationsN/A | Result date |
| Arenatext factuality | Categorycreative writing | Rank87 | Rating / score1,376.23 | Votes7,119 | ObservationsN/A | Result date |
| Arenatext factuality | Categorymulti turn | Rank87 | Rating / score1,419.40 | Votes7,724 | ObservationsN/A | Result date |
| Arenatext | Categoryfrench | Rank88 | Rating / score1,419.73 | Votes1,958 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry writing and literature and language | Rank89 | Rating / score1,393.67 | Votes11,661 | ObservationsN/A | Result date |
Step 3.5 Flash Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $0.10 input, $0.30 output per 1M
- Official provider
- StepFun (Global)
- Lowest third-party
- From $0.09 input, $0.30 output per 1M via Vercel AI Gateway
- Tracked offerings
- 13
Sort by
Provider | Provider model ID | Region | Input / 1M | Output / 1M | Context | Updated |
|---|
| ProviderNvidia | Provider model IDstepfun-ai/step-3.5-flash | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context256K | Updated |
| ProviderVercel AI Gateway | Provider model IDstepfun/step-3.5-flash | Regionglobal | Input / 1M$0.09 | Output / 1M$0.30 | Context262.1K | Updated |
| ProviderNanoGPT | Provider model IDstepfun-ai/step-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context262.1K | Updated |
| ProviderHugging Face | Provider model IDstepfun-ai/Step-3.5-Flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context262.1K | Updated |
| ProviderOpenRouter | Provider model IDstepfun/step-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context262.1K | Updated |
| ProviderStepFun (Global) | Provider model IDstep-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context256K | Updated |
| ProviderStepFun (China) | Provider model IDstep-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context256K | Updated |
| ProviderZenMux | Provider model IDstepfun/step-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context256K | Updated |
| ProviderEmpirioLabs AI | Provider model IDstep-3-5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context256K | Updated |
| ProviderKilo Gateway | Provider model IDstepfun/step-3.5-flash | Regionglobal | Input / 1M$0.10 | Output / 1M$0.30 | Context262.1K | Updated |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Step 3.5 Flash Runtime Performance
Provider-specific output speed and catalog latency for Step-3.5-Flash. Runtime does not affect the capability score.
Sort by
Provider | Output Speed | Catalog Latency | Max Input | Max Output | Updated |
|---|
| ProviderStepFun | Output Speed21.16 tok/s | Catalog Latency1.90 s | Max Input65.5K | Max Output8.2K | Updated |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Step 3.5 Flash Specifications
Technical details for the model's default version.
- Version
- Step-3.5-Flash
- Released
- Feb 2, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 196B
- Context window
- 65.5K
- Max output
- 8.2K
- Inputs
- image, text
- Outputs
- text
- Open weights
- Yes
- License
- Apache 2.0
Step 3.5 Flash Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionStep-3.5-Flash | Released | LLMBoard59.42 | Parameters196B | Context65.5K | Max output8.2K | Open weightsYes | LicenseApache 2.0 |
Models similar to Step 3.5 Flash
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
ST
Step3 VL 10B
StepFun
ME
LongCat Flash Thinking
Meituan
BA
ERNIE 5.0
Baidu
AC
Qwen3.6 27B
Alibaba Cloud / Qwen Team
OP
GPT-5.4-mini
OpenAI
MA
Kimi K2 Thinking
Moonshot AI
What is Step 3.5 Flash?
Key information about Step 3.5 Flash and its available data.
5 Flash is a large language model built on StepFun's Step3 architecture. It is designed for text tasks with low latency and efficient token usage.
Data as of 2026-09-08.
FAQ
Common questions about Step 3.5 Flash.
When was Step 3.5 Flash released?
Step 3.5 Flash's default version was released on Feb 2, 2026.
How much does Step 3.5 Flash cost?
Step 3.5 Flash's official API price is $0.10 per million input tokens and $0.30 per million output tokens via StepFun (Global). The lowest tracked third-party offer starts at $0.09 input and $0.30 output via Vercel AI Gateway.
Who created Step 3.5 Flash?
Step 3.5 Flash was created by StepFun.
What is the context window for Step 3.5 Flash?
The default version has a 65.5K token context window.
Is Step 3.5 Flash open weight?
Yes. The default version is marked as open weight under Apache 2.0.