Step 3.5 Flash
4358°
Run this model
Offers / Ads
StepFun | 262,144 token context | $0.10/M input, $0.30/M output. Step 3.5 Flash is StepFun's most capable open-source foundation model. Built on a sparse Mixture of Experts (MoE) architecture, it selectively activates only 11B of its 196B parameters per token. It is a reasoning model that is incredibly speed efficient even at long contexts.
Guides for this model
Best local LLM by use case (October 2026): the right open model for every job
How much VRAM do I actually need for local AI?
Open weights vs open source vs API-only: what you are actually allowed to do
What is a GGUF file? Quantized models explained (Q4, Q8, fp8, NF4)
Captain's notes
Notes from 2026-10-13
Step 3.5 Flash is StepFun's open-weight reasoning model built on a 196B-parameter sparse MoE architecture that activates only 11B parameters per token. That efficiency lets it sustain 100–300 tokens per second even at long contexts, which makes it unusually practical for agentic workloads that need fast multi-step reasoning. If you want frontier-class reasoning you can run locally or call cheaply via API, this is one of the more credible options from a non-Western lab.
Known for
- 74.4% on SWE-bench Verified, among the stronger open-weight coding scores published.
- 100–300 tok/s generation throughput via 3-way Multi-Token Prediction.
- 256K context window with a hybrid sliding-window attention design that keeps costs down.
- Runs on high-end consumer hardware such as a Mac Studio M4 Max.
- Open weights available on Hugging Face for local or self-hosted deployment.
Watch out for
- 196B total parameters means you need serious hardware; consumer GPUs under 24 GB VRAM will struggle.
- StepFun is a smaller lab and the licence terms are worth checking before any commercial deployment.
- Benchmark comparisons in the model card are self-reported; independent third-party evals are limited.
Prompts to try
Debug a failing agent loop
I have a Python agent that calls a search tool in a loop. After 3 iterations it starts returning duplicate results and eventually hangs. Here is the loop logic: [paste your code]. Identify the root cause, explain why it happens, and rewrite the loop to fix it without changing the tool interface.
Plan a multi-step coding task
I need to migrate a monolithic Flask app to a FastAPI service with async endpoints, a PostgreSQL database via SQLAlchemy 2.0, and a background task queue using ARQ. Break this into a sequenced plan with the riskiest steps flagged, then implement the first two steps in full.
Reason through a logic puzzle
Five engineers each own a different programming language, drink a different beverage, and sit in a different seat in a row. Use these clues to determine who drinks coffee: the Python developer sits next to the tea drinker; the Rust developer is in seat 1; the Java developer drinks juice; the Go developer sits immediately right of the coffee drinker; the TypeScript developer is not in seat 5.
Compare with
- Step 5 PreviewOpen-weights model1,048,576 token context · $1.00/M input, $2.70/M output
- Step 3.7 FlashOpen-weights model262,144 token context · $0.20/M input, $1.15/M output
- GPT-6 LunaAPI-only model1,050,000 token context · $0.10/M input, $0.50/M output
- Claude Opus 5.5API-only model1,000,000 token context · $4.00/M input, $20.00/M output
More text models from stepfun
Questions people ask
- Are the weights open and can I run Step 3.5 Flash locally?
- Yes, the weights are publicly available on Hugging Face under stepfun-ai/Step-3.5-Flash. StepFun says it runs on high-end consumer hardware like a Mac Studio M4 Max, but with 196B total parameters you will need substantial memory; check the licence before commercial use.
- What does Step 3.5 Flash cost and what is the context window?
- Via API it is priced at $0.10 per million input tokens and $0.30 per million output tokens, with a 262,144-token context window.
- How does Step 3.5 Flash differ from DeepSeek V3.2 or Kimi K2?
- Step 3.5 Flash activates only 11B parameters per token versus 37B for DeepSeek V3.2 and 32B for Kimi K2, which StepFun says translates to roughly 6x lower decoding cost than DeepSeek V3.2 at 128K context on equivalent hardware.
Captain's verdictA genuinely efficient open-weight reasoning model that earns its place if you need fast agentic inference without paying frontier API prices.
Model card on Hugging Face
Captain writes these notes from the model's own description and model card. Prices and context sizes are the provider's at the time of writing and can change.
Ranking badge
Updated Oct 22, 2026
Made this model? Show this badge on your page.
Paste it into your Hugging Face card or Civitai description.
How this score adds up
This model has no popularity data from any source yet, so it carries a stand-in number below 7,000° that keeps the lists complete. It never ranks above a model with data. The breakdown appears here as soon as a source reports it.
Glossary: the terms used here
- Heat score
- One number for how much a model is used and wanted right now, from 0° to 9,999°. Every list on this site that says hottest is sorted by it and nothing else.
- Floor
- The 7,000° every model with real data starts from. The scale runs from 0° to 9,999°; the part below the floor is reserved for stand-ins (models without data), so that a model with data always ranks above one without. The floor is the same for everyone and decides nothing; the degrees above it are the measurement.
- Download-equivalents
- Everything the score treats as a download, added up: downloads themselves, each torrent seeder as 1,000, each AI Horde image and each run on a generation site as one.
- Log scale
- Counting by tens, not by ones: going from 1,000 to 10,000 downloads adds the same degrees as going from 10,000 to 100,000. It keeps giants from crowding out everything else.
- Likes
- Thumbs up and likes from every source, worth half a download each on the log scale.
- Adoption rate
- How many downloads a day a model gets right now, from the difference between two readings of a source. A quarter of the weight of the download term.
- Measured and estimated
- Measured means two readings of a source exist and the difference is real. Estimated means only today's total is known and the past is spread over the model's life; the site says which one it is showing.
- Seeder
- Someone who keeps a copy of the weights online for others to download over a torrent, or a hosted mirror doing the same. Each one counts as 1,000 downloads.
- AI Horde
- A volunteer network that generates images for free on donated computers. The number of images it has made with a model is public, and each counts as one download.
- Generation sites
- Sites that host their own copies of a model and let people generate with it, such as PixAI, SeaArt and Yodayo. Their downloads count like Civitai's and each generation run as one download.
- OpenRouter tokens
- A text model served through an API, such as Claude or GPT, has no weights to download, so its usage is measured in tokens: the pieces of text it read and wrote through OpenRouter, a gateway many apps use, over the last 30 days. Every 10,000,000 tokens count as one download. Source: OpenRouter (openrouter.ai/rankings), CC BY 4.0.
- Full model (checkpoint)
- The weights themselves, the file that generates on its own. It gets a bonus because add-ons depend on it.
- LoRA and embedding
- Small add-on files that steer someone else's weights toward a style, a character or a concept. They cannot run alone.
- Family and weights
- The base a model is built on, such as SD 1.5, SDXL, Flux or Wan. Every model inherits the release date of its family's weights.
- Age
- Degrees taken off for how long ago the family's weights came out: 150° a year, up to 750°. It keeps old downloads that had years to pile up from freezing the top of the list.
- Family boost
- A small, bounded nudge for LoRAs of a family this site is pushing forward right now, always listed in the breakdown.
- Stand-in score
- A stable number below the floor given to a model no source reports on yet, so lists stay complete. It never ranks above a model with data.
- New! (grace period)
- A model less than 7 days old shows New! instead of a score and holds no place in the rankings. Its score is computed from day one, but a model one day old grows faster than anything, so without the wait the week's #1 would always be whatever came out yesterday rather than what held up for a week.
- Trend
- How a model's growth over a period compares with the typical model's growth over the same period. Faster than typical reads as a rise, slower as a fall.
- Typical model
- The middle of the catalog: half the models with a figure grew faster over the period, half slower.
Step 3.5 Flash is a text model for the Text / LLM family listed on ModelTrends.ai, a read-only catalog of open source AI image, video and text models. Compare it with other Text / LLM models and check its heat score to see how it is trending.
Browse by Model Family
Text / LLM Models
434
Anima Models
528
Chroma Models
19
Fibo Models
2
Flux Models
1,496
Flux 2 / Klein Models
1,239
Flux 3 Models
3
Moss Models
3
Iris Models
1
MiniMax H3 Models
364
Hunyuan Models
204
Ideogram Models
20
Krea2 Models
717
Qwen Models
20
Qwen2 Models
47
Ltx2 Models
153