# Mistral Small 4 API

Mistral Small 4 (API version mistral-small-2603, released 16 March 2026) is Mistral's hybrid model unifying instruct, reasoning and coding in one 119B-parameter mixture-of-experts, with text and image input and a 256K-token context window; list price $0.15 in / $0.60 out per 1M tokens, provider list as at 1 November 2026.

- Provider: Mistral
- Model ID: mistral-small-4
- Type: Chat
- Region: Global (provider global routing)
- Context window: 256,000 tokens
- Max output: not listed
- Modalities: text, vision, tools, reasoning
- Endpoints: OpenAI Chat Completions https://api.seedrise.ai/v1
- Prices checked: 1 November 2026

## Provider list price

Provider list price as at 1 November 2026, per 1M tokens. AUD at a reference rate of 1 USD = 1.53 AUD; invoices use the rate on the day.

| Token type | USD | AUD |
|---|---:|---:|
| Input | 0.15 | 0.23 |
| Output | 0.60 | 0.92 |
| Cached input | Rates on request | Rates on request |

Your Seedrise rate is quoted per account against this list price. Request a rate card: https://seedrise.ai/contact?model=mistral-small-4

Global: runs on the provider's global routing; no Australian-region option as at 1 November 2026. Cached input is not listed by the provider and is quoted per account.

## Strengths

- Configurable reasoning: none for quick replies, high for step-by-step reasoning
- Text and image input
- Open weights under the Apache 2.0 licence
- Function calling and structured outputs

## Typical use cases

- Agentic coding and codebase exploration
- Document parsing and visual analysis
- General chat assistants at low cost
- Math and research tasks with reasoning on

## Quickstart

A first request with curl. Replace sk-... with your Seedrise key.

```bash
curl https://api.seedrise.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{"model":"mistral-small-4","messages":[{"role":"user","content":"Hello"}]}'
```

OpenAI SDK, Python:

```python
from openai import OpenAI

client = OpenAI(base_url="https://api.seedrise.ai/v1", api_key="sk-...")

r = client.chat.completions.create(
    model="mistral-small-4",
    messages=[{"role": "user", "content": "Summarise this contract in three bullets."}],
)
print(r.choices[0].message.content)
```

## Where it runs

Global: Served on the provider's global routing.

## Related models

- [Mistral Large 4](https://seedrise.ai/models/mistral-large-4): US$0.70 input · US$2.10 output per 1M tokens, Global
- [Qwen3.8 Flash](https://seedrise.ai/models/qwen3.8-flash): US$0.15 input · US$0.47 output per 1M tokens, SG
- [Claude Haiku 5.5](https://seedrise.ai/models/claude-haiku-5-5): US$0.10 input · US$0.50 output per 1M tokens, AU
- [GPT-6 Luna](https://seedrise.ai/models/gpt-6-luna): US$0.10 input · US$0.50 output per 1M tokens, Global

## FAQ

### What is the model ID for Mistral Small 4?

Use mistral-small-4 as the model field, with no vendor prefix, on the OpenAI-compatible endpoint at https://api.seedrise.ai/v1. GET /v1/models returns the IDs your key can call.

### How does the Seedrise rate for Mistral Small 4 compare to the list price?

Mistral's list price as at 1 November 2026 is US$0.15 per 1M input tokens and US$0.60 per 1M output tokens. Seedrise quotes rates per account against the provider's published list price for the same routing. Volume and committed tiers sit below list.

### What is the context window and max output of Mistral Small 4?

Mistral Small 4 has a context window of 256,000 tokens, per the provider documentation checked on 1 November 2026. Maximum output is not listed on our rate card; ask us for the current limit.

### Can Mistral Small 4 run in Australia?

Not at present. Mistral Small 4 is served on the provider's global routing. Models marked AU in the model library can be pinned to Sydney or Melbourne on request.

### How do I switch reasoning on or off for Mistral Small 4?

Use the reasoning_effort parameter: none gives fast everyday replies and high gives step-by-step reasoning, per Mistral.

### Can I self-host Mistral Small 4 later?

Yes. Mistral publishes Small 4 under the Apache 2.0 licence, so you can evaluate it on Seedrise first and run the weights yourself later.

Model names are trademarks of their respective providers.

Source: https://seedrise.ai/models/mistral-small-4