Baltor Get started

Models

Step 3.5 Flash

By StepFun. 199.4 billion parameters, a context window of 262,144 tokens and the licence apache-2.0.

Facts

Released
2026-02-01 the day the repository was first published on Hugging Face Hugging Face, read
Licence
apache-2.0 the licence the model card declares Hugging Face, read
Open weights
Yes the weights are published in this Hugging Face repository Hugging Face, read
Parameters
199.4 billion counted from the safetensors weight files Hugging Face, read
Active parameters
Unknown
Knowledge cutoff
Unknown
Context window
  • 262,144 tokens the maximum position embeddings in the model configuration model configuration, read
  • 262,144 tokens the context length OpenRouter lists OpenRouter, read
Longest output
  • 65,536 tokens the largest output OpenRouter's first provider allows OpenRouter, read
Tool calling
  • Yes OpenRouter lists tools among the supported parameters OpenRouter, read
Structured output
  • No OpenRouter lists no structured output parameter OpenRouter, read
Reasoning controls
  • Yes OpenRouter lists reasoning controls OpenRouter, read
Inputs and outputs
text in, text out Hugging Face, read
Good for
  • Reasoning: OpenRouter lists reasoning controls for it OpenRouter, read

Prices

Prices in US dollars per million tokens, as each source lists them
ProviderInputOutputContextAs ofSource
Nvidia direct0 USD0 USD256,000 older than 30 days; check the providermodels.dev, read
Hugging Face direct0.100 USD0.300 USD262,144 older than 30 days; check the providermodels.dev, read
NanoGPT direct0.100 USD0.300 USD262,144 older than 30 days; check the providermodels.dev, read
SiliconFlow direct0.100 USD0.300 USD262,000 older than 30 days; check the providermodels.dev, read
SiliconFlow through OpenRouter0.100 USD0.300 USD262,144OpenRouter endpoints, read
SiliconFlow (China) direct0.100 USD0.300 USD262,000 older than 30 days; check the providermodels.dev, read

Run it on your own hardware

Weights are the sizes of the files a source lists, or an estimate from the parameter count where none does. Memory is an estimate: weights plus KV cache plus 512 MiB and 5 percent of the weights for runtime buffers. Check it against your hardware.

QuantizationWeightsMemory at 8,192 tokensMemory at 32,768 tokens
Q4_K111 GiB129 GiB165 GiB

Start it with a local runtime

  • Ollama
    ollama run hf.co/ggml-org/Step-3.5-Flash-GGUF:Q4_K
  • LM Studio
    lms get ggml-org/Step-3.5-Flash-GGUF
lms server start
  • llama.cpp server
    llama-server -hf ggml-org/Step-3.5-Flash-GGUF:Q4_K --jinja
  • vLLM
    vllm serve stepfun-ai/Step-3.5-Flash
  • SGLang
    sglang serve --model-path stepfun-ai/Step-3.5-Flash --port 30000

Use it from your harness

Set up for Hugging Face with the model stepfun-ai/Step-3.5-Flash. Each endpoint page has the same setup for its own address.

OpenCode

Put this in opencode.json in your project folder:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "huggingface": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Hugging Face",
      "options": {
        "baseURL": "https://router.huggingface.co/v1",
        "apiKey": "{env:HF_TOKEN}"
      },
      "models": {
        "stepfun-ai/Step-3.5-Flash": {
          "name": "Step 3.5 Flash",
          "limit": {
            "context": 262144,
            "output": 65536
          }
        }
      }
    }
  }
}
  • OpenCode reads any OpenAI-compatible address through the @ai-sdk/openai-compatible package, and an address that speaks the Responses API through @ai-sdk/openai.

From OpenCode documentation, read .

Pi

Put this in ~/.pi/agent/models.json:

{
  "providers": {
    "huggingface": {
      "baseUrl": "https://router.huggingface.co/v1",
      "api": "openai-completions",
      "apiKey": "$HF_TOKEN",
      "models": [
        {
          "id": "stepfun-ai/Step-3.5-Flash"
        }
      ]
    }
  }
}
  • The apiKey field can name an environment variable as $NAME.

From Pi documentation, read .

Codex

Codex speaks only the Responses API, and Hugging Face documents no Responses address. A gateway that offers one can sit in between.

  • Codex speaks the Responses API only: responses is the one supported wire API of a custom provider. Ollama and LM Studio are built in and start with --oss.

From Codex documentation, read .

Claude Code

Claude Code sends Anthropic Messages requests, and Hugging Face documents no such address. A gateway that translates to that API can sit in between.

  • Claude Code sends Anthropic Messages requests to ANTHROPIC_BASE_URL. Anthropic says it does not support routing Claude Code to models other than Claude through any gateway, so some features may not work with another model.

From Claude Code documentation, read .

Other ways to reach it

Published results

These are results other people published. Baltor did not run them and does not rank models by them.