vs
Nebius vs StepFun
A Shanghai lab selling its own efficient multimodal models against a European cloud that hosts many labs' open weights with in-region placement.
By The Subconscious Team · Updated
Nebius vs StepFun: key differences
StepFun is a model maker. Its workhorse Step 3.7 Flash is a 198B mixture-of-experts vision-language model with 11B active parameters, 256K context and an Apache 2.0 license, priced at $0.20 in and $1.15 out on StepFun's own API. Nebius is a host. It serves 60+ open models from several labs on Token Factory and rents GPUs, with EU or US placement for dedicated endpoints. The comparison is really between buying from the lab and running open weights on a neutral cloud.
Data location drives much of that decision. StepFun's first-party inference is China-hosted, and it has thin Western distribution and support. Nebius targets European enterprises that need workloads kept in-region. Because Step 3.7 Flash ships under Apache 2.0 and runs on vLLM and SGLang, a team could self-host it on Nebius GPUs, though it is not listed in Nebius's catalog here. StepFun's direct API suits cost-sensitive vision and video understanding where data location is flexible. Nebius suits teams that want a broader catalog, SLAs and control over where tokens are processed.
What Nebius and StepFun do
Nebius
Nebius is an Amsterdam-headquartered AI cloud and the strongest European alternative to the US hyperscalers. It sells raw NVIDIA GPU compute, from H100s at $2.15 an hour preemptible up to GB300 NVL72 racks, and it has begun adding Vera Rubin. Hyperscale buyers back it: a Microsoft capacity deal worth about $17.4B in September 2025, then a Meta agreement worth up to about $27B in March 2026.
Example models: DeepSeek V3, GPT-OSS
Full Nebius profileStepFun
StepFun is a Shanghai AI lab known for efficient multimodal models, with a mix of proprietary API models and open-weight releases. Its current 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 3.7 Flash, Step3
Full StepFun profileShould you choose Nebius or StepFun?
Nebius
Choose Nebius for
- Open-model inference kept in the EU or US
- A multi-lab catalog under one account and SLA
- Self-hosting permissively licensed weights on rented GPUs
StepFun
Choose StepFun for
- Low-cost image and video understanding with 256K context
- Direct access to Step models from the lab that trains them
- Small-active-parameter models for cheap self-hosting
Nebius vs StepFun at a glance
| Attribute | Nebius | StepFun |
|---|---|---|
| Model access | Open weights, 60+ models | Open (Apache 2.0) and API models |
| Flagship models | DeepSeek, Qwen, GLM, Kimi, GPT-OSS | Step 3.7 Flash, Step3 |
| Speed | Among top hosts on throughput | ~128 tok/s on Step 3.7 Flash |
| Price | From $0.06 per 1M input | $0.20 in, $1.15 out (Step 3.7 Flash) |
| Customization | Serve uploaded fine-tunes | Open weights to fine-tune |
| Deployment | Token Factory, dedicated, raw GPUs | First-party API, OpenRouter |
| Long context | Varies by model | 256K |
Frequently asked questions
What is the difference between Nebius and StepFun?
A Shanghai lab selling its own efficient multimodal models against a European cloud that hosts many labs' open weights with in-region placement.
When should I choose Nebius over StepFun?
Open-model inference kept in the EU or US; A multi-lab catalog under one account and SLA; Self-hosting permissively licensed weights on rented GPUs.
When should I choose StepFun over Nebius?
Low-cost image and video understanding with 256K context; Direct access to Step models from the lab that trains them; Small-active-parameter models for cheap self-hosting.
Is Nebius or StepFun cheaper?
Nebius: From $0.06 per 1M input. StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). The cheaper choice depends on the model and workload.
Which has more context, Nebius or StepFun?
Nebius: Varies by model. StepFun: 256K.
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.