Solar Mini 4
Solar Mini 4 is Upstage's compact, cost-efficient language model: a 35B-parameter Mixture-of-Experts with just 3B active parameters and a 524K-token context window. At $0.05 per million input tokens it is the cheapest model in this group, built for agentic use where response cost dominates.
Technical Specifications
Live data fetched from the OpenRouter model registry — always up to date.
What is Upstage: Solar Mini 4?
Solar Mini 4 is built on a clear premise: a 35B-parameter Mixture-of-Experts that activates only 3B parameters per token, paired with a 524,288-token context window. That combination is unusual - long context usually arrives with a long bill. Upstage targets agentic use cases where the same large context is sent repeatedly, and at $0.05 per million input tokens the arithmetic works.
Features & Capabilities
3B Active of 35B
A sparse MoE that keeps compute per token very low.
$0.05 per Million
The cheapest model in this group by some distance.
524K Context
Half a million tokens at a price that permits repetition.
Agent-Tuned
Built for agentic workflows using AI Agents.
Solar Line
The compact tier of Upstage established Solar family.
Korean and English
Strong on both, which matters for Korean-language work.
Who Is Upstage: Solar Mini 4 For?
Repetitive Agent Calls
Sending the same large context over and over cheaply.
High-Volume Classification
Running a simple decision across huge volumes.
Korean-Language Work
Content and analysis where Korean support matters.
Cost-Ceiling Pipelines
Keeping unit economics viable on thin margins.
Upstage: Solar Mini 4 Pricing
- $0.05 per million input tokens, $0.20 per million output tokens
- Cheapest model in this batch by a wide margin
- Run it from your Hoodgen free credits, no credit card required
- Higher allowances on Hoodgen AI paid plans
What Makes Upstage: Solar Mini 4 Different?
Solar Mini 4 is a lesson in what sparse architectures make possible. Thirty-five billion total parameters but only three billion active per token means the model is genuinely small to run, which is how Upstage can offer a 524,288-token context at $0.05 per million input tokens. Context length and cost are normally locked together; this decouples them.
The target workload is agentic repetition. An agent that sends the same long system context on every step pays for that context every time, so input price dominates the bill. Cutting it to five cents per million changes how many steps you can afford, and how much context you can leave in place.
It is not a frontier model and does not pretend to be. Three billion active parameters will not match Opus 5.5 on hard reasoning. But for routing, classification, extraction and the cheaper steps of a larger workflow it is close to free, and Hoodgen letting you pair it with a strong planner is the pattern that gets the most out of it.
Upstage: Solar Mini 4 Performance & Pricing
Solar Mini 4 is the cheapest model in this batch and pairs that with a 524K-token context window.
What Can You Do With Upstage: Solar Mini 4?
Solar Mini 4 fits wherever cost per call is the binding constraint and the task is well defined. It is the model to pair with a strong planner when you need many cheap steps rather than one clever one.
Best Use Cases
- Routing and classification at volume
- Cheap steps inside a larger agent workflow
- Korean-language content tasks
- Extraction into structured fields
- Pipelines where input cost dominates
Upstage: Solar Mini 4 Privacy & Data Policy
Solar Mini 4 is developed by Upstage and served through OpenRouter. Prompts and completions are processed under Upstage terms.
As with any third-party model, avoid pasting credentials or personal information. Hoodgen AI keeps stronger models on the same account so the few hard steps can be escalated without changing tools.
Upstage: Solar Mini 4 vs Other Models
Frequently Asked Questions
Half a million tokens for five cents a million
Chat with Solar Mini 4 on Hoodgen - sign in and start in seconds.
Chat with Upstage: Solar Mini 4