vs
Fireworks AI vs StepFun
StepFun is a Shanghai lab with cheap Apache 2.0 multimodal models. Fireworks is a host for 400+ open models, with training and compliance built in.
By The Subconscious Team · Updated
Fireworks AI vs StepFun: key differences
StepFun makes models; Fireworks serves them. StepFun's workhorse, Step 3.7 Flash, is a 198B mixture-of-experts vision-language model with only 11B active parameters, 256K context and selectable reasoning levels, priced at $0.20 in and $1.15 out on StepFun's own API. It ships under Apache 2.0 and runs on vLLM and SGLang. Fireworks hosts 400+ open models across text, vision, audio and embeddings, with a stack that posts 167 to 174 tokens per second on DeepSeek V4 Pro and keeps that model's full 1M window.
StepFun's first-party inference is China-hosted, with thin Western distribution and support, and its models trail the frontier on hard multimodal reasoning. Fireworks offers SOC 2, HIPAA and ISO, cloud marketplace billing and managed SFT, DPO and RL. Step 3.7 Flash is a strong pick for cheap image and video understanding, and because the weights are open, a team can self-host it or reach it through OpenRouter. Teams that need a certified host, larger models or custom training fit Fireworks better. Check whether Fireworks lists Step models before planning to run them there.
What Fireworks AI and StepFun do
Fireworks AI
Fireworks AI was founded in 2022 by former Meta PyTorch engineers led by CEO Lin Qiao, and it sells speed on open models. Its custom serving stack has posted 167 to 174 tokens per second on DeepSeek V4 Pro in third-party measurements, several times what most GPU peers hit on the same model. The catalog holds 400+ models across text, vision, audio and embeddings, served through an OpenAI-compatible API. In July 2026 it raised a $1.505B Series D at a $17.5B valuation, with a reported $1B+ run rate and 40T+ tokens a day.
Example models: DeepSeek V4 Pro, Kimi K3
Full Fireworks AI 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 Fireworks AI or StepFun?
Fireworks AI
Choose Fireworks AI for
- Certified hosting with SOC 2, HIPAA and ISO
- Larger open models like DeepSeek V4 Pro at 1M context
- Managed RL and SFT fine-tuning
StepFun
Choose StepFun for
- Cheap vision and video understanding in agents
- Self-hosting a small-active-parameter model under Apache 2.0
- Long-context multimodal work at $0.20 per million input
Fireworks AI vs StepFun at a glance
| Attribute | Fireworks AI | StepFun |
|---|---|---|
| Model access | Open weights | Open (Apache 2.0) and API models |
| Flagship models | DeepSeek V4 Pro, Kimi K3 | Step 3.7 Flash, Step3 |
| Speed | 167–174 tok/s on DeepSeek V4 Pro | ~128 tok/s on Step 3.7 Flash |
| Price | Fine-tunes served at base price | $0.20 in, $1.15 out (Step 3.7 Flash) |
| Customization | SFT, DPO, RFT; Training API | Open weights to fine-tune |
| Deployment | Serverless, dedicated GPUs | First-party API, OpenRouter |
| Long context | Full 1M on DeepSeek V4 Pro | 256K |
Frequently asked questions
What is the difference between Fireworks AI and StepFun?
StepFun is a Shanghai lab with cheap Apache 2.0 multimodal models. Fireworks is a host for 400+ open models, with training and compliance built in.
When should I choose Fireworks AI over StepFun?
Certified hosting with SOC 2, HIPAA and ISO; Larger open models like DeepSeek V4 Pro at 1M context; Managed RL and SFT fine-tuning.
When should I choose StepFun over Fireworks AI?
Cheap vision and video understanding in agents; Self-hosting a small-active-parameter model under Apache 2.0; Long-context multimodal work at $0.20 per million input.
Is Fireworks AI or StepFun cheaper?
Fireworks AI: Fine-tunes served at base price. StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). The cheaper choice depends on the model and workload.
Which has more context, Fireworks AI or StepFun?
Fireworks AI: Full 1M on DeepSeek V4 Pro. 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.