vs
Meta vs StepFun
Two labs mixing closed APIs and open weights. Meta offers Muse Spark with 1M context and cheap media add-ons; StepFun offers cheaper Apache 2.0 multimodal models.
By The Subconscious Team · Updated
Meta vs StepFun: key differences
Both labs play both sides. Meta's closed Muse Spark 1.3 has a 1M token context at $1.25 in and $4.25 out, and its open-weight Muse Glimmer, distilled from Spark, runs on vLLM, SGLang, llama.cpp or ExecuTorch. StepFun's Step 3.7 Flash is a 198B mixture-of-experts vision-language model with 11B active parameters, 256K context and Apache 2.0 weights, priced at $0.20 in and $1.15 out on StepFun's API. StepFun is cheaper per token. Meta offers four times the context on its hosted model.
Beyond price, distribution and jurisdiction matter. StepFun's first-party inference is China-hosted with thin Western distribution and support, and it trails frontier models on hard multimodal reasoning. Meta's API is US-based but new, and its cheapest tier trades your data for the discount. Meta adds Muse Image at $0.01 per image and transcription at $0.18 per hour. Cost-sensitive vision and video understanding, or cheap self-hosting under Apache 2.0, fits StepFun. Agentic coding and computer use at 1M context fits Meta.
What Meta and StepFun do
Meta
Meta has moved from open Llama releases toward its own closed API. Meta Superintelligence Labs builds the Muse family, and in July 2026 Meta launched the Meta Model API with Muse Spark 1.1, a multimodal reasoning model aimed at agentic coding, tool use and computer use. At Meta Connect in September, Meta made the API generally available globally. Muse Spark 1.3 has a 1M token context and costs $1.25 in and $4.25 out per million tokens, with cached input at $0.15. Muse Wonder, released October 12, 2026, sits above it for longer, harder agent work at $4 in and $20 out, also with 1M context. The endpoint speaks OpenAI Chat Completions, Anthropic Messages and a stateful agentic format.
Example models: Muse Wonder, Muse Spark 1.3, Muse Spark Open
Full Meta profileStepFun
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. It costs $1.00 in and $2.70 out per million tokens, and its weights followed on Hugging Face in mid-October under the custom StepFun Model License 1.0, which allows commercial use with restrictions. 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 profileShould you choose Meta or StepFun?
Meta
Choose Meta for
- Agentic coding and computer use at 1M context
- A US-hosted API with OpenAI and Anthropic formats
- Image generation and transcription on the same key
StepFun
Choose StepFun for
- Low-cost image and video understanding
- Apache 2.0 weights with only 11B active parameters
- Budget multimodal agents that fit in 256K
Meta vs StepFun at a glance
| Attribute | Meta | StepFun |
|---|---|---|
| Model access | Closed API; open Muse Spark Open and Glimmer | Open weights (Apache 2.0 or StepFun license) and API models |
| Flagship models | Muse Wonder, Muse Spark 1.3, Muse Spark Open | Step 5 Preview, Step 3.7 Flash |
| Speed | ~145–233 tok/s on Muse Spark 1.3 | ~128 tok/s on Step 3.7 Flash |
| Price | $1.25–$4 in, $4.25–$20 out; Contributor tier cheaper | $0.20 in, $1.15 out (Step 3.7 Flash) |
| Customization | Open Muse weights to fine-tune | Open weights to fine-tune |
| Deployment | Meta Model API, Oracle Cloud | First-party API, OpenRouter |
| Long context | 1M | Up to 1M (Step 5 Preview); 256K on Step 3.7 Flash |
Frequently asked questions
What is the difference between Meta and StepFun?
Two labs mixing closed APIs and open weights. Meta offers Muse Spark with 1M context and cheap media add-ons; StepFun offers cheaper Apache 2.0 multimodal models.
When should I choose Meta over StepFun?
Agentic coding and computer use at 1M context; A US-hosted API with OpenAI and Anthropic formats; Image generation and transcription on the same key.
When should I choose StepFun over Meta?
Low-cost image and video understanding; Apache 2.0 weights with only 11B active parameters; Budget multimodal agents that fit in 256K.
Is Meta or StepFun cheaper?
Meta: $1.25–$4 in, $4.25–$20 out; Contributor tier cheaper. StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). The cheaper choice depends on the model and workload.
Which has more context, Meta or StepFun?
Meta: 1M. StepFun: Up to 1M (Step 5 Preview); 256K on Step 3.7 Flash.
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.