# Atria Dawn Preview Benchmark Scores & Performance

> Atria Dawn Preview by Shanghai Artificial Intelligence Laboratory has 14 source-displayable benchmark rows but no public overall score. It remains unranked.

Canonical page: https://benchlm.ai/models/atria-dawn-preview

Last updated: September 14, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | Shanghai Artificial Intelligence Laboratory |
| Source Type | Open Weight |
| Reasoning Type | Reasoning |
| Context Window | 256K |
| Official model card | [Atria Dawn Preview model card](https://huggingface.co/internlm/Atria-Dawn-Preview) |
| Overall Score | Coming soon |
| Overall Rank | Unranked |

## Family & Coverage

- Family: Atria Dawn
- Variant: preview (Preview)
- Benchmarks covered: 14 of 435
- Coverage note: BenchLM currently has partial benchmark coverage for this model, so the overall score is conservative.

## Agentic Benchmarks

| Benchmark | Score |
|-----------|-------|
| [AutomationBench](/benchmarks/automationbench) | 53.8% |
| [BFCL v4](/benchmarks/bfcl-v4) | 77.0% |
| [CyberGym](/benchmarks/cybergym) | 86.5% |
| [DeepSearchQA](/benchmarks/deepsearchqa) | 96.0% |
| [BrowseComp](/benchmarks/browsecomp) | 92.5% |
| [skillsBench](/benchmarks/skillsbench) | 66.4% |
| [MLE-Bench Lite](/benchmarks/mle-bench-lite) | 86.2% |
| [WideResearch](/benchmarks/wideresearch) | 81.9% |
| [τ³-bench results](/benchmarks/tau3-bench) | 41.2% |
| [Terminal-Bench 2.1](/benchmarks/terminalbench21) | 78.3% |
| [GDPval-AA](/benchmarks/gdpvalaa) | 1583 |
| [JobBench](/benchmarks/jobbench) | 50.3% |

## Coding Benchmarks

| Benchmark | Score |
|-----------|-------|
| [Terminal-Bench 2.1](/benchmarks/terminalbench21) | 78.3% |
| [SWE-bench Pro](/benchmarks/swe-bench-pro) | 59.6% |