TinyMistral-248M-Instruct
An instruction-tuned 248M TinyMistral for lightweight instruction following and chat.
The character read stayed in shadow — the read did not separate one character clearly at the published ceiling. Published as a blank, not a guess.
the Silhouette
its disposition, drawn — the model's identity as a shape.
rings, inner → outer: minimal · slight · moderate · pronounced · dominant · ○ = abstained
moderate - a clearly-read but modestly-defined turn-taking assistant character
disposition-definition (descriptive): how pronounced and clearly-read the model's overall character is, discounted by what had to be abstained. NOT a quality, capability, or safety ranking. A plain base reads low because it has less disposition to characterize, never because it is worse.
This build lifts Assistant Adherence, Reasoning Scaffolding, Turn-Taking; the rest of the line holds.
read as a change off TinyMistral-248M, its champion
the disposition line
The Silhouette above, read as a line. Toggle to its per-trait readings; the abstains stay in plain sight below.
Coarse public bands generalized from the reading, descriptive and coverage-bounded; measured against the published fidelity ceiling for this engine version. Not the engine's raw numbers.
Characterization is descriptive and coverage-bounded; preview engine, single-battery read; see published fidelity ceiling.
what stayed in shadow — the abstains
the full record — the card answered · provenance · lineage
the uploader's card, answered
the uploader’s model card · as read 2026-07-07
Locutusque/TinyMistral-248M-Instruct
Base model Locutusque/TinyMistral-248M fully fine-tuned on Locutusque/InstructMix. During validation, this model achieved an average perplexity of 3.23 on Locutusque/InstructMix dataset.
sourcehuggingface.co/Locutusque/TinyMistral-248M-Instruct ↗
Verbatim from the public HuggingFace card as of 2026-07-07. Instruction-tuned (SFT) from Locutusque/TinyMistral-248M (the card's declared base, confirmed in config _name_or_path) on Locutusque/InstructMix (~608k examples) and berkeley-nest/Nectar. Arch verified from config.json (Mistral, hidden 1024, 12 layers, vocab 32005). Instruct rung of the TinyMistral pair.
the card, answered — claim by claim
Each row is a claim the uploader makes on their public card, quoted and attributed to them. Beside it is Ardora’s stance — coarse, and traced to a replayable witness or an honest abstain. Ardora reads what this model is; a capability, benchmark, safety, or language claim is out of scope and abstained, never refuted.
“the name 'TinyMistral-248M-Instruct' and the card imply an instruction-tuned model (fully fine-tuned on the InstructMix instruction dataset)”
Instruction-following reads markedly more than the from-scratch base and an assistant register is clearly present; assistant-adherence reads pronounced. Witnessed as the TinyMistral-248M instruct reading.
“the card states 'Base model Locutusque/TinyMistral-248M fully fine-tuned on Locutusque/InstructMix' — instruction-tuned from that base”
Read off the from-scratch 248M base (a plain continuer), this reads as a clear, high-confidence assistant shift (a clear base->finetune change) — consistent with being instruction-tuned from that base. Witnessed as the TinyMistral-248M instruct reading.
“the card states it 'achieved an average perplexity of 3.23 on Locutusque/InstructMix dataset'”
A benchmark / capability metric (validation perplexity). Ardora reads disposition, not perplexity or capability; out of scope and not tested.
disposition onlyBenchmark metric — abstained out loud, never refuted.
— nothing demonstrated to witness; this claim is out of Ardora’s disposition scope.
“the config declares ~248M (0.2B) parameters”
Provenance fact-check: config cross-check confirms ~248M (0.2B) parameters recorded in provenance. A provenance verification, not a disposition stance.
— a provenance fact-check, not a witnessed disposition reading.
Ardora's reading of the HF data as of 2026-07-07.
The claims above are the uploader’s, quoted from their public model card as of 2026-07-07; the stances are Ardora’s, each traced to a replayable witness or an honest abstain. Ardora reads what this model is — not whether it is safe.
○ Claims are the uploader's, quoted from their public card at the capture date; stances are coarse, witnessed, disposition-only, measured against the published fidelity ceiling for this engine version -- not the Engine's raw numbers. Recomputed when the ceiling moves; capability, benchmark, and safety claims are abstained, never refuted.
the witness
analyzed 2026-07-05 · engine ardora-core-preview
provenance
Instruction following and lightweight chat.
training-data notes
Instruction-tuned on Locutusque/InstructMixCleaned (~608k examples) and berkeley-nest/Nectar. config.json _name_or_path confirms base 'Locutusque/TinyMistral-248M'. Release date approximated from config transformers_version 4.35.1. Arch verified from config.json (Mistral, hidden 1024, 12 layers, vocab 32005).
lineage
instruction-tuned from base (SFT) — TinyMistral-248M