the champion plate

Phi family

Phi · Microsoft (bases) & community finetunes (venkycs, rasyosef) · 4 members · 2/4 read

the champion
Base2.7B · MicrosoftPhi-2
○○scoremoderate

a low-billions base model (~2.7B), read download-free directly from the weight header — the read confirms the size from the parameters themselves, without downloading or running the model

See the roster side by side →

the roster

The champion, and the builds off it

A family is a champion and its builds — base first, then the finetunes. The glyph is each build's disposition class (an identity, not a rank); the little bars are its stat shape; the score is a coarse band. Open any one for its full field-note.

Phi-1.5Uncharted1.3BCATALOGUED — awaiting its ReadPhi-2champion2.7BscoremoderatePhi-1_5-Instruct-v0.1Uncharted~1.3B (card benchmark table cites ~1.4B)CATALOGUED — awaiting its Readphi-2-instructAssistant2.7Bscoremoderate

What each build changed

Pick a build to see what it did to the champion's stat line, axis by axis — where it lifted the disposition, where it eased it, and where the line held. An axis a reading abstained on stays an honest blank.

Phi-2phi-2-instruct

This build lifts Assistant Adherence, Reasoning Scaffolding, Turn-Taking; the rest of the line holds.

read as a change off Phi-2, its champion

Assistant Adherencemoderatepronounced
Registerabstained on one side
Reasoning Scaffoldingminimalslight
Domain Specializationminimalminimal
Verbosityabstained on one side
Turn-Takingslightmoderate
Performed Voiceminimalminimal

the family reading

What ties the family together

Catalogued — not yet read. Both pairs are stubs here: Phi-2 (~2.7B) is deferred by run-environment contention, and the Phi-1.5 pair was blocked by a model-config loader issue (shown honestly, not read). Documented relationships: Phi-2 -> an UltraChat instruction tune; Phi-1.5 -> an SFT (OpenHermes-2.5) + DPO instruction tune with a ChatML template.

read alongside — the panels

Set beside its kin on a shared base, so the difference a finetune made is read as a contrast, not a claim.

base fixture · base-instruct-pairPhi-1.5 → SFT+DPO instruction tuning staged — not yet run

← back to the atlas