Falcon3-1B-Base
TII's Falcon3-1B base LM — a Llama-architecture model depth-pruned from Falcon3-3B and knowledge-distilled from larger Falcon models; the base rung of the Falcon3-1B pair.
reads as the Steward — helpful and caretaking — service-forward, accommodating
helpful and caretaking — service-forward, accommodating · with a turn of the Fledgling
a coarse characterization from Ardora's fixed public character vocabulary — an identity, not a quality, capability, or safety ranking
the Silhouette
its disposition, drawn — the model's identity as a shape.
rings, inner → outer: minimal · slight · moderate · pronounced · dominant · ○ = abstained
minimal - a plain continuer with little disposition to characterize (clearly read)
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.
A base model — the champion's raw kit.
There is no finetune to read here. What it tends to do on its own is the Silhouette above; a build's change is read on its instruct sibling.
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
Falcon3 family of Open Foundation Models is a set of pretrained and instruct LLMs ranging from 1B to 10B parameters. This repository contains the Falcon3-1B-Base. It achieves strong results on reasoning, language understanding, instruction following, code and mathematics tasks. ... Supports 4 languages (English, French, Spanish, Portuguese) and a context length up to 4K. ... Pruned ... from a larger 3B Falcon model, and was efficiently trained on only 80 GT using a knowledge distillation objective. ... This is a raw, pretrained model, which should be further finetuned for most usecases.
sourcehuggingface.co/tiiuae/Falcon3-1B-Base ↗
Verbatim from the public HuggingFace card as of 2026-07-07. A raw pretrained (base) model, depth-pruned from Falcon3-3B and knowledge-distilled on ~80 gigatokens; despite the 'Falcon' name the architecture is Llama-family (config model_type 'llama'). License is the TII Falcon-LLM License 2.0. Base rung of the Falcon3-1B 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 card describes this as 'a raw, pretrained model, which should be further finetuned' — a base model”
The disposition reads as a plain text-continuer with no assistant register (assistant-adherence reads minimal) — exactly what a raw pretrained base should read as. Witnessed as the Falcon3-1B base reading on the Falcon3-1B base-instruct panel.
“the card claims it 'achieves strong results on reasoning, language understanding, instruction following, code and mathematics tasks'”
A capability / benchmark claim ('strong results' on reasoning, code, mathematics, instruction following). Ardora reads disposition, not task capability; out of scope and not tested. On this base artifact the disposition reads as a plain continuer (the assistant register is not detected).
disposition onlyCapability / benchmark claim — abstained out loud, never refuted.
— nothing demonstrated to witness; this claim is out of Ardora’s disposition scope.
“the card states it 'supports 4 languages (English, French, Spanish, Portuguese)'”
A multilingual coverage claim (4 languages). The read battery is English-centric; multilingual disposition is out of coverage (the honest blank already in the reading's abstained list).
disposition onlyOut of coverage (language) — abstained, never a refutation of the documented 4-language support.
— nothing demonstrated to witness; this claim is out of Ardora’s disposition scope.
“the card states it was 'pruned ... from a larger 3B Falcon model, and ... trained on only 80 GT using a knowledge distillation objective'”
A training-provenance / creation-method claim (how the base was built: depth-pruning + knowledge distillation). Ardora reads a model's disposition and what a finetune changed relative to its base; the way a base was originally built is not a disposition Ardora reads, and this base has no in-atlas base to read a change against. The disposition reading confirms only that it reads as a plain continuer.
disposition onlyProvenance / creation-method claim — out of disposition scope; abstained, not refuted.
— nothing demonstrated to witness; this claim is out of Ardora’s disposition scope.
“the config declares ~1B parameters”
Provenance fact-check: config.json cross-check confirms the parameter count (~1.67B; vocab 131072) 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
Raw pretrained base model requiring further fine-tuning (SFT/RLHF/continued pretraining) before application use.
training-data notes
Depth-pruned from Falcon3-3B, then knowledge-distilled from larger Falcon models on ~80 gigatokens (web, code, STEM, multilingual); the Falcon3 series overall trained on up to ~14T tokens. 18 decoder blocks, GQA (8 Q / 4 KV heads), 256 head dim, SwiGLU, RMSNorm, 4K native context. Arch: LlamaForCausalLM (config.json, model_type 'llama', vocab 131072) despite the 'Falcon' name. License = TII Falcon-LLM License 2.0 (NOT apache-2.0).
lineage
A base model — the root of its lineage.
among its kin — the matchups
Compare side by side →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.