Add Orion Flagship Mini benchmark results
Add Refract-Labs/Orion-Flagship-Mini to BananaMind Base Bench 1.1.
Official complete benchmark run:
- Overall Elo: 1030
- Raw accuracy: 54.86% (192/350)
- Language Completion Elo: 1225
- Commonsense Elo: 996
- World Knowledge Elo: 1006
- Context Tracking Elo: 923
- Quantitative Elo: 858
- Logical Reasoning Elo: 1036
- Code Completion Elo: 1201
- Parameters: ~219.1M
Model:
https://huggingface.co/Refract-Labs/Orion-Flagship-Mini
The run completed all 350 benchmark cases with no context truncations.
Adds a new 'Refract AI Labs' org and an 'Orion-Flagship-Mini' model entry to script.js, but the new org record is missing the logo field every other org has and reuses an existing org's chart color.
Thanks for the PR! The model entry itself follows the existing shape and the numbers look plausible, but a few fixes are needed before merging:
Missing
logofield: every org inORGShas alogoURL, but the newrefractlabsentry omits it. Depending on howscript.jsrenders orgs this will show a broken image or throw. Please add the org's logo (e.g. the avatar URL from https://huggingface.co/Refract-Labs).Duplicate chart color:
chartColor: "rgba(34, 211, 238, 0.70)"/chartBorder: "#22d3ee"is already used by thecnwplayerorg, so the two would be indistinguishable in charts. Please pick an unused color.Formatting: the diff re-indents the existing
bananamind:key and the firstMODELSentry's opening{to 6 spaces, inconsistent with the rest of the file. Please keep the original 2-space indentation and match the style of the neighboring entries.
Once the logo is added and the color/indentation are fixed, this should be good to go.
Automated review by BananaMindBot using z-ai/glm-5.3-flash. A maintainer can override this.
SHUTUP IF IT WORKS IT WORKS
SHUTUP IF IT WORKS IT WORKS
its fine im adding those fixes rn