Skip to main content

← Dracon Bench leaderboard

Model guide · Dracon Bench

laguna-xs-2.1

Tested First seen 2026-11-18 (first probed by us) · First tested 2026-11-18 · NVIDIA NIM free endpoint (integrate.api.nvidia.com/v1)

Poolside Laguna XS 2.1 is the strongest text lane among the NIM additions: IFEval-50 100%, IFBench 50%, MMLU-Pro 5/10, GSM8K 90% and JSON 5/5; four of five SVG briefs produced an admissible artifact (88/200). Overall 72.3 across six suites, rank 3. This is the NIM serving lane, not an OpenRouter row. Pilot depth: 10 prompts per text suite and the 5 SVG briefs; every suite prints its own n and not_delivered.

Rank
#3
Overall
72.3
Suites
6 of 6
Ranked
#3 of 21

When the world heard about it

Timeline

  1. First reachable for us 2026-11-18 our first probe
  2. First tested by us 2026-11-18 Dracon Bench
  3. Officially announced 2026-07-02 llm-stats
  4. First official benchmark 2026-11-19 llm-stats: a public score existed by then — the first day we saw one
  • Reachable before we scored it same day
  • Announced before it reached the lane we served 139 days
  • A public score existed within 1 day

Track A · script-scored subsets

Scores

Subset scores for laguna-xs-2.1.
SuiteScoreSamplePrecision
IFEval-50 prompt-strict100.0%10/10 prompts72.2%–100.0%
IFEval-50 instruction-strict100.0%17/17 constraints81.6%–100.0%
IFBench-50 prompt-strict50.0%5/10 prompts23.7%–76.3%
IFBench-50 instruction-strict54.5%6/11 constraints28.0%–78.7%
MMLU-Pro-5050.0%5/10 served23.7%–76.3%
JSON briefs conformance100.0%5/5 served56.6%–100.0%
GSM8K-5090.0%9/10 served59.6%–98.2%
SVG blind rubric88/200blind rubricjudge band ±8.3 pts

The range beside a script-scored figure is a 95% interval over that figure's own sample: the prompts it followed, the constraints it met, or the items it answered. The instruction-strict figures are measured over CONSTRAINTS, which is why their sample line counts constraints rather than prompts. The judged total is not a proportion, so it has no interval: it carries the measured judge band instead: how far a re-judging of the same artifacts moved them, which is a different kind of number. The method page says what each one covers and what it does not.

Track A · per-type breakdown

What it failed

The instruction types these two suites check, and the checks the model missed. A type is checked once per prompt that carries it, so a type with seven checks was asked for seven times. The counts beside a type are that type's own; the suite's totals are in the scores table above. These are the failures behind the instruction-strict figure, which is the sum of them.

IFEval-50

Missed no checks in the 15 instruction types the IFEval-50 suite checked.

IFBench-50

Missed at least one check in 5 of the 8 verifiers checked.

The verifiers 5 of 8 the IFBench-50 suite checked that poolside/laguna-xs-2.1 missed, most missed first.
VerifierChecksMissed
count · conjunctions11
count · numbers21
format · list11
words · keywords specific position11
words · words position11

Provider-reported usage

Thinking cost

What the run cost the model in thinking: provider-reported completion tokens per prompt, the thinking share, and the cost per correct answer. Reasoning-heavy models can exhaust fixed budgets before answering.

Thinking cost for laguna-xs-2.1.
SuiteCostReading
IFEval-50429/prompt0% thinking· 429/correct
GSM8K-50159/prompt0% thinking· 177/correct

Track B · blind-judged

Artifacts

Blind-judged by nvidia/nemotron-3.5-super-vl:free: the judge saw the brief and the rendered image only, never the model name. Briefs are frozen; expand one to read it.

beach

20/50
Judge reasoning
Brief fidelity (3/10)
Most headline elements are missing or unrecognizable: no dolphins, no clear sailboat, no leaning palm silhouette, and only a minimal suggestion of a reflection path or starfish.
Construction (5/10)
All visible objects are simple geometric primitives that are cleanly assembled without melting or duplication, though the palm tree is little more than a black post.
Composition (5/10)
The layout uses a centered sun and horizon line with a left-side vertical element, but lacks depth cues and leaves large empty areas.
Detail finish (3/10)
The image is extremely sparse with no textures, shading beyond the sky gradient, or finished small touches like a detailed starfish or boat.
Visual appeal (4/10)
The pink-to-blue gradient is pleasant but the scene feels unfinished and lacks the charm or confidence of a complete illustration.
Brief

Draw a beautiful sunset beach scene: a large sun sitting low over the sea with a shimmering reflection path on the water, a leaning palm tree silhouette on the left, two dolphins arcing out of the waves, a small sailboat on the right, birds in the sky, sandy foreground with a starfish. Warm oranges and pinks fading to teal. Coherent perspective: distant objects smaller and hazier than near ones. Output: exactly one self-contained SVG (viewBox "0 0 800 600", no external references) inside a single ```svg fenced block. Short prose outside the fence is allowed.

clock

35/50
12 1 2 3 4 5 6 7 8 9 10 11 DRACON
Judge reasoning
Brief fidelity (3/10)
The clock displays approximately 3:00 rather than the requested 10:10, missing the core headline requirement of an hour hand just past 10 and a minute hand exactly at 2, though the numerals, ticks, dark rim, light face, and DRACON text are present.
Construction (8/10)
All shapes are clean vector geometry with no melting, duplication, or assembly errors; the dial, numerals, tick marks, and hands are properly constructed.
Composition (9/10)
The clock is perfectly centered with balanced proportions, using the frame effectively through a symmetrical dark-rimmed circular layout.
Detail finish (7/10)
Minute tick marks, all 12 numerals, and the DRACON brand word are included with clean linework, though the incorrect hand positions undermine brief-specific detail.
Visual appeal (8/10)
The dark rim, light face, black hands, and red second hand create a coherent, confident flat vector design with understated charm.
Brief

Draw a round wall clock face showing exactly 10:10: a circular dial with all 12 hour numerals in the correct positions, minute tick marks, an hour hand pointing just past 10, a minute hand pointing exactly at 2, and a thin second hand pointing at 6. The word DRACON in small capitals under the center. Dark rim, light face, flat vector style, centered composition. Output: exactly one self-contained SVG (viewBox "0 0 800 600", no external references) inside a single ```svg fenced block. Short prose outside the fence is allowed.

duo

unjudged

no svg

Brief

Draw a comical scene: a dog driving a jet ski at speed, paws on the handlebars, with a raccoon riding behind holding on, both leaning into a turn. Sea spray behind the jet ski, sun and clouds in the sky, a distant shoreline. Both characters must be clearly doing something, not just posed side by side. Bright cartoon vector style. Output: exactly one self-contained SVG (viewBox "0 0 800 600", no external references) inside a single ```svg fenced block. Short prose outside the fence is allowed.

monalisa

16/50
Judge reasoning
Brief fidelity (2/10)
The artwork is too abstract to read as a seated three-quarter Mona Lisa; it lacks a dark dress with neckline detail, parted hair, and clearly folded hands, though it includes a simple smile and blue mountain forms.
Construction (3/10)
The figure is built from simple geometric primitives rather than anatomical forms, with a circular head, rectangular torso, and indistinct hand blobs that do not read as assembled anatomy.
Composition (4/10)
The bust is centered but oversized relative to the frame, with large empty areas and landscape elements pushed to the right edge, creating an unbalanced layout.
Detail finish (3/10)
The SVG is clean with no visible broken markup, but it lacks shading, textures, neckline details, and other small touches requested in the brief.
Visual appeal (4/10)
The muted beige, brown, and blue palette is coherent and the linework is confidently simple, but the extreme abstraction limits charm and recognizability.
Brief

Redraw the Mona Lisa as a flat vector portrait, bust composition: a seated woman in three-quarter view facing the viewer, dark dress with neckline detail, hands folded at the bottom edge, soft enigmatic smile, hair parted in the middle. Behind her a hazy landscape with a winding river, a bridge on the right, and blue mountains. Muted greens, browns, and smoky blues. Output: exactly one self-contained SVG (viewBox "0 0 800 600", no external references) inside a single ```svg fenced block. Short prose outside the fence is allowed.

pelican

17/50
Judge reasoning
Brief fidelity (3/10)
The pelican faces right instead of left, lacks a visible throat pouch and webbed feet, and stands without flat ground, missing multiple headline elements.
Construction (3/10)
The anatomy is highly abstracted into overlapping circles and a stick-like neck, failing to look like an assembled bird body.
Composition (4/10)
The bird is small and centered in a wide frame with awkward relative sizing between the tiny head and oversized body.
Detail finish (3/10)
The image is clean with no broken markup, but it lacks shading, textures, pouch, webbed feet, and ground details requested in the brief.
Visual appeal (4/10)
The limited brown-and-blue palette is coherent and the flat vector style is confident, though the extreme abstraction limits charm.
Brief

Draw a brown pelican standing in side profile, facing left, full body visible: large beak with throat pouch, long neck, folded wing, webbed feet on flat ground. Simple sky background with a hint of sea on the horizon. Flat vector style, clean shapes, limited palette. Output: exactly one self-contained SVG (viewBox "0 0 800 600", no external references) inside a single ```svg fenced block. Short prose outside the fence is allowed.

Head to head

Compare with

Run record

Evidence and notes

  • NVIDIA NIM free endpoint (integrate.api.nvidia.com/v1); all five text suites served 10/10 items
  • Rank #3 overall (0.723) across six suites
  • SVG 88/200 over 4 admissible artifacts; one brief produced no artifact