# Cohere vs StepFun

> Two labs with Apache 2.0 flagships. StepFun's Step 3.7 Flash is cheaper and multimodal; Cohere adds retrieval models and Western enterprise deployment.

Canonical: https://www.subconscious.dev/compare/cohere-vs-stepfun · By The Subconscious Team · Updated September 30, 2026

## How they compare

The flagships look alike on paper. Cohere's Command A+ is a 218B mixture-of-experts model with 25B active, Apache 2.0, 128K context, and reasoning, vision, translation and tool use in one set of weights. StepFun's Step 3.7 Flash is a 198B MoE vision-language model with 11B active, Apache 2.0, 256K context and selectable reasoning levels. StepFun publishes $0.20 in and $1.15 out per million on its own API, while Cohere does not publish Command A+ prices and lists Command A at $2.50 in and $10 out. StepFun also reads video, and its speed is listed around 128 tokens per second against Cohere's reported 375 on Command A+ in 4-bit form.

Distribution and support separate them more than weights. StepFun's first-party inference is China-hosted with thin Western support, and outside its own API it reaches developers mainly through OpenRouter or self-hosting on vLLM and SGLang. Cohere sells through Bedrock, Azure, OCI and its own API, and it deploys into any VPC or on-prem with fine-tuning. Cohere adds Embed 4, Rerank 4, Aya and Transcribe. StepFun is cheaper for vision and video understanding in cost-sensitive agents, and its small active parameter count keeps self-hosting light. Western regulated buyers will usually find Cohere easier to procure.

## What each one does

### Cohere

Cohere is a Toronto-based lab that sells models and platforms to banks, governments and large enterprises rather than consumers. Its generative line is the Command family. Command A+, released May 20, 2026, is a 218B-parameter mixture-of-experts model with 25B active, published under Apache 2.0 with a 128K context window, and it combines reasoning, vision, translation and tool use in one set of weights. Command A has a 256K window and lists at $2.50 in and $10 out per million tokens, while Command R7B costs $0.0375 in. June 2026 added North Mini Code, a 30B Apache 2.0 coding model, and the lineup also includes Aya multilingual models and Transcribe for speech.

### StepFun

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.

## Which is best, and when

### Choose Cohere for

- Western enterprise procurement and support
- RAG with first-party embeddings and rerank
- On-prem fine-tuning under compliance rules

### Choose StepFun for

- Low-cost image and video understanding
- Self-hosting a small-active MoE model
- Long 256K context at low token prices

## At a glance

| Attribute | Cohere | StepFun |
|---|---|---|
| Model access | Closed, plus open Command A+ | Open (Apache 2.0) and API models |
| Flagship models | Command A+, Command A, Embed 4, Rerank 4 | Step 3.7 Flash, Step3 |
| Speed | 375 tok/s on Command A+ W4A4, per Cohere | ~128 tok/s on Step 3.7 Flash |
| Price | $0.0375–$2.50 in, $0.15–$10 out per 1M | $0.20 in, $1.15 out (Step 3.7 Flash) |
| Customization | Enterprise fine-tuning, incl. private | Open weights to fine-tune |
| Deployment | API, Bedrock, Azure, OCI, VPC, on-prem | First-party API, OpenRouter |
| Long context | 256K on Command A; 128K on A+ | 256K |

## FAQ

### What is the difference between Cohere and StepFun?

Two labs with Apache 2.0 flagships. StepFun's Step 3.7 Flash is cheaper and multimodal; Cohere adds retrieval models and Western enterprise deployment.

### When should I choose Cohere over StepFun?

Western enterprise procurement and support; RAG with first-party embeddings and rerank; On-prem fine-tuning under compliance rules.

### When should I choose StepFun over Cohere?

Low-cost image and video understanding; Self-hosting a small-active MoE model; Long 256K context at low token prices.

### Is Cohere or StepFun cheaper?

Cohere: $0.0375–$2.50 in, $0.15–$10 out per 1M. StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). The cheaper choice depends on the model and workload.

### Which has more context, Cohere or StepFun?

Cohere: 256K on Command A; 128K on A+. StepFun: 256K.

## Running long-horizon agents?

If your agents run past 200K tokens, compare both against Subconscious: [Subconscious vs Cohere](https://www.subconscious.dev/compare/subconscious-vs-cohere.md), [Subconscious vs StepFun](https://www.subconscious.dev/compare/subconscious-vs-stepfun.md).

Full profiles: [Cohere](https://www.subconscious.dev/providers/cohere.md), [StepFun](https://www.subconscious.dev/providers/stepfun.md).