U
Pareto
Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks.
Specifications
| Lab | Unbiased |
|---|---|
| Context window | 262.1K tokens |
| Input price | $2.50/1M |
| Output price | $7.50/1M |
| Release | Sep 2026 |
Strengths
- Blended multi-model answers on every request
- 74 on DeepSWE, tying larger frontier models
- Never switches models mid-conversation, preserving prompt cache
- Frontier-level output at $2.50/$7.50 per million tokens
Weaknesses
- New lab with no long-term track record
- Trails leaders on HLE and Terminal-Bench
- Composite approach adds latency vs single-model calls
Best for
- Agentic coding workflows
- Research tasks needing cross-checked answers
- Cost-sensitive frontier-level work
In Depth: Pareto
Summary
Pareto is an AI model from Unbiased.
Released in Sep 2026. It takes text and image input and produces text output, with a context window of 262.1K tokens. Input pricing is $2.50 per 1M tokens and output is $7.50 per 1M tokens on OpenRouter.