vs
StepFun vs Boundless
StepFun is a Shanghai lab with its own multimodal models and Apache 2.0 weights. Boundless serves open models like Kimi K3 plus GPT and Claude from the US or Europe.
By The Subconscious Team · Updated
StepFun vs Boundless: key differences
StepFun's Step 5 Preview is a 600B mixture-of-experts model with 27B active, 1M context and image input, and Artificial Analysis scores it level with Kimi K3. Step 3.7 Flash ships under Apache 2.0 with 256K context and costs $0.20 in and $1.15 out per million tokens on StepFun's own API, which is hosted in China. Boundless serves Kimi K3 at $2.30 in and $11.40 out, along with GLM-5.3, DeepSeek V4.1 Flash and other open models, most with 1M context, plus resold GPT and Claude models. It runs requests in the US or Europe with zero data retention, and Vercel's AI Gateway measured it at 138 tokens per second on Kimi K3.
Pick StepFun for low-cost image and video understanding, or Apache 2.0 weights to run yourself. Pick Boundless for processing in the US or Europe, GPT and Claude on the same key as open models, or a reserved deployment its engineers tune to contract targets.
What StepFun and Boundless do
StepFun
StepFun is a Shanghai AI lab known for efficient multimodal models, with a mix of proprietary API models and open-weight releases. In September 2026 it launched Step 5 Preview, its new flagship for agentic work: a 600B-total, 27B-active mixture-of-experts model with 1M context and text and image input, which Artificial Analysis scores level with Kimi K3. Its cheaper workhorse, Step 3.7 Flash, came out in May 2026 as a 198B mixture-of-experts vision-language model with only 11B active parameters. It has 256K context, selectable reasoning levels, tool use and structured outputs, and it ships under Apache 2.0. StepFun's own API prices it at $0.20 in and $1.15 out per million tokens, and OpenRouter carries it too.
Example models: Step 5 Preview, Step 3.7 Flash
Full StepFun profileBoundless
Boundless launched in September 2026 as an inference partner for AI-native startups. It sells two things. The Model API serves pre-tuned models on shared capacity, billed per token from prepaid credits with a budget per API key. Tailored inference reserves capacity for one customer, and Boundless engineers pick the model on the customer’s own evals, tune precision, batching, caching, speculative decoding and parallelism, then run the deployment against uptime, throughput and latency targets written into the contract. A free two-week evaluation tests those targets on real traffic against the customer’s current provider first.
Example models: GLM-5.3, DeepSeek V4.1 Flash, Kimi K3
Full Boundless profileShould you choose StepFun or Boundless?
StepFun vs Boundless at a glance
| Attribute | StepFun | Boundless |
|---|---|---|
| Model access | Open (Apache 2.0) and API models | Open weights, plus closed GPT and Claude |
| Flagship models | Step 5 Preview, Step 3.7 Flash | GLM-5.3, DeepSeek V4.1 Flash, Kimi K3 |
| Speed | ~128 tok/s on Step 3.7 Flash | 342 tok/s on DeepSeek V4.1 Flash 1017 (Vercel) |
| Price | $0.20 in, $1.15 out (Step 3.7 Flash) | $0.04–$2.30 in, $0.21–$11.40 out per 1M (open) |
| Customization | Open weights to fine-tune | Serving-stack tuning, no fine-tuning |
| Deployment | First-party API, OpenRouter | Serverless API, reserved tailored capacity |
| Long context | Up to 1M (Step 5 Preview); 256K on Step 3.7 Flash | 1M context on most models |
Frequently asked questions
What is the difference between StepFun and Boundless?
StepFun is a Shanghai lab with its own multimodal models and Apache 2.0 weights. Boundless serves open models like Kimi K3 plus GPT and Claude from the US or Europe.
When should I choose StepFun over Boundless?
Low-cost vision-language models; Self-hosting Step 3.7 Flash; Agentic work on Step 5 Preview.
When should I choose Boundless over StepFun?
US or Europe processing; Kimi K3 and GLM-5.3 at 1M context; Engineer-run reserved capacity.
Is StepFun or Boundless cheaper?
StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). Boundless: $0.04–$2.30 in, $0.21–$11.40 out per 1M (open). The cheaper choice depends on the model and workload.
Which has more context, StepFun or Boundless?
StepFun: Up to 1M (Step 5 Preview); 256K on Step 3.7 Flash. Boundless: 1M context on most models.
Related comparisons
Running long-horizon agents?
If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.