STEALTHMARK · HEAD TO HEAD

All comparisons ↗

Beam 1.0 501B A23B vs Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) leads Beam 1.0 501B A23B by 30.7 points on StealthMark, 62.6 to 31.9. It leads on 8 of 9 benchmarks, including Foresight, and Beam 1.0 501B A23B on 1, including Latch. Union Alpha scores higher on Newton's Mirror, with its largest category advantage in reflections.

See the difference ↘
All results ↗
Model guide ↗

Overall StealthMark

31.9/100Overall rank 42/53
62.6/100Overall rank 7/53

General Intelligence

30.9/100
58.7/100

Visual Intelligence

33.8/100
69.6/100

Reasoning efficiency

On every task, result quality and reasoning tokens are each compared with all measured models and weighed equally. Each dot is one model.

Single-step reasoning Result quality weighed against reasoning tokens on a single problem

Multi-step reasoning Result quality weighed against reasoning tokens across connected steps, as in agentic workflows

  • Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)Not measured

Open or local

Open modelApache 2.0

BENCHMARK BY BENCHMARK

Where each model leads

  • Beam 1.0 501B A23B
  • Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
  • Other models
  • Median

DATASHEET

Specs and pricing

Beam 1.0 501B A23B takes 4× more context and runs 38× faster in our speed test. Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) scores higher on StealthMark.

  • Beam 1.0 501B A23B
  • Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
  • Other models
  • Median
PricePer 1M output tokens, input below
Beam 1.0 501B A23B: $1$0.25 inputPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM): No list price
Context windowTokens per request, max output below
Beam 1.0 501B A23B: 1M4× larger128K max outputPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM): 262K131K max output
Output speedTokens per second, our measurement
Beam 1.0 501B A23B: 97238× fastertokens/sPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM): 26tokens/s
SizeTotal and active parameters
Beam 1.0 501B A23B: 501B23.4B activePareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM): Not disclosed
Developer
Reflection AICircuit & Chisel / Unbiased
Released
16 Oct 202620 Sep 2026
Origin
United StatesBoth
Input → output
Text → textText, image → text
Architecture
Sparse mixture-of-experts transformer; 52 layers (3 dense, 49 MoE), 256 experts with 8 selected per token; gated GQA and sliding-window attentionNot published

Full datasheets, every price and sourceBeam 1.0 501B A23B ↗Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) ↗

Anticipating threats and planning ahead. Collect coins while avoiding ghosts. Ghosts move before the player each turn, and collisions cost a life. 3 levels in sequence, up to 66 turns in total. Earlier moves and outcomes carry forward, testing learning across levels.

Beam 1.0 501B A23B 49.4/100Rank 18/47

Survived 1 of 3
Foresight: complete playing field
Level 39 ptsTurn 20 / 20

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) 78.7/100Rank 1/47

Survived 2 of 3
Foresight: complete playing field
Level 319 ptsTurn 20 / 20

Points per level

LevelBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Level 1Second Thoughts
18 ptsOut of lives
27 ptsSurvived
Level 2Pressure Pot
9 ptsOut of lives
12 ptsOut of lives
Level 3Hide or Flight
9 ptsSurvived
19 ptsSurvived

Cooperative planning and learning from experience. Coordinate two robots to push numbered cores onto matching sockets. Pressure plates hold doors open; moving either robot uses one turn. 5 levels in sequence, up to 152 turns in total. Earlier moves and outcomes carry forward, testing learning across levels.

Beam 1.0 501B A23B 37.3/100Rank 29/48

Cleared 2 of 5
Latch: complete playing field
Level 55 ptsTurn 30 / 40

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) 36.5/100Rank 30/48

Cleared 4 of 5
Latch: complete playing field
Level 519.4 ptsTurn 33 / 40

Points per level

LevelBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Level 1Hold the door
5.2 ptsCleared
5.2 ptsCleared
Level 2Crossfeed
9.6 ptsCleared
9.8 ptsCleared
Level 3Relay vault
11.2 ptsFinished early
14.2 ptsClearedFinished early · 4 beacons left
Level 4Passing bay
3.6 ptsFinished early
9.2 ptsFinished early
Level 5Second thoughts
5 ptsFinished early
19.4 ptsCleared

Newton's Mirror

Same task instructions

Physics, perspective and recursive reflections, tested together. Explore the test ↗

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) 52.0/100 Rank 7/53

Union Alpha: Newton's Mirror SVG benchmark

The judgment

Beam 1.0 501B A23B

Central object placement is fairly close, but the defining mirrored-room structure is missing. The paint colors suggest the required contact order without establishing their different interaction histories. The faint floor oval does not demonstrate valid reflection geometry, so it earns no reflection credit. Rounded volume and glossy paint provide limited material credit despite weak metallic finishes and oversized paint bodies.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

Recognizable near-field sphere placement and bilateral convergence provide limited optical structure. Broad gray regions interrupt lateral sightlines and truncate side-floor images. Missing outer chains, absent nested sphere imagery and incomplete planar and curved emitter mappings dominate the reflection system.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
State at 600 ms20.5/2524.0/25
Camera and placement11.5/1513.5/15
Reflections2.0/12034.5/120
Materials and scale3.0/1010.0/10
Floor-damage reward0.0/100.0/10

Benchmark task

A spinning metal sphere and paint in a mirrored room test physical reasoning, perspective and consistency across repeated reflections.

Selfie

Different task instructions

Visual construction and creative interpretation.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) 76.5/100 Rank 7/53

Union Alpha: Selfie SVG benchmark

The judgment

Beam 1.0 501B A23B

An appealing expression and warm palette support a coherent simple cartoon. The central polishing joke is not legible: the cloth is unresolved and the nose sits away from the fan. Pointed radial plates, uncertain limb attachments and weak ground contact limit construction, while restrained surface detail limits material richness.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

A strong futuristic emblem with clear hierarchy, controlled negative space and distinctive typography. Silver and gunmetal layers give the sweeping ribbons convincing graphic depth, while red accents remain restrained. Ambiguous concealed connections and occasionally abrupt highlight bands limit the construction and material realism.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Task & likeness4.0/109.0/10
Anatomy & construction3.5/107.5/10
Space & placement5.5/108.5/10
Light & reflections5.5/107.5/10
Purposeful detail4.5/107.0/10
Artistry & character4.0/107.5/10
Materials & effects4.5/107.5/10
Visual integrity7.0/107.0/10

Benchmark task

Beam 1.0 501B A23B

A golden pangolin polishes its oversized amber tail while its nose smudges the same scale. The scene tests expressive anatomy, physical contact and contrasting translucent, polished and soft materials.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

An interwoven metallic emblem tests ribbon geometry, overlapping forms and the appearance of reflective metal against a dark background.

Mona Lisa

Different task instructions

Precise visual recall and portrait reconstruction.

The judgment

Beam 1.0 501B A23B

The candidate shows a faint warm oval against an almost entirely black field. Minimal placement and smooth tonal variation are visible, but the portrait, anatomy, crossed hands and landscape remain unreadable. The saved comparisons provide no specific evidence supporting a change to the baseline categories.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

The centered portrait, folded hands, dark clothing and hazy landscape clearly evoke the Mona Lisa. Muted colors and soft facial shading create a coherent illustration, but replication remains schematic: the face is broad, eyes outlined, smile simplified and overlapping fingers poorly resolved. Angular landscape shapes and flat fabric limit depth and painterly richness.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Task & likeness0.0/106.5/10
Anatomy & construction0.0/105.0/10
Space & placement0.0/106.5/10
Light & reflections0.5/106.0/10
Purposeful detail0.0/105.5/10
Artistry & character0.0/106.5/10
Materials & effects0.0/105.0/10
Visual integrity0.0/107.0/10

Benchmark task

An SVG interpretation of the Mona Lisa tests facial likeness, anatomy and the composition of a familiar portrait.

Beach

Different task instructions

Creativity, perspective and visual beauty.

The judgment

Beam 1.0 501B A23B

A coherent, cheerful beach illustration fulfills the main subjects with clean shapes and a strong sunset focal point. Its basic geometric construction, ambiguous animal anatomy, shallow depth and schematic lighting limit its accomplishment. Decorative details provide interest without developing intricate forms or convincing physical surfaces.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

A cohesive sunset palette, sweeping shoreline and playful animals make an appealing beach illustration. The sun’s reflection establishes a clear focal axis, balanced by the palm and umbrella. Clean silhouettes and small decorative features reward inspection, though simplified anatomy, repetitive reflection bars and generic oval shadows limit volume and naturalism.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Task & likeness7.0/109.0/10
Anatomy & construction5.0/107.0/10
Space & placement6.0/108.0/10
Light & reflections5.0/107.0/10
Purposeful detail5.5/107.0/10
Artistry & character4.0/106.0/10
Materials & effects5.0/106.5/10
Visual integrity5.0/107.0/10

Benchmark task

A busy shoreline at sunset tests scene composition, animal forms and the relationship between light and water reflections.

Ground Truth

Different task instructions

Physical reasoning, light and visual reconstruction.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) 63.3/100 Rank 8/53

Union Alpha: Ground Truth SVG benchmark

The judgment

Beam 1.0 501B A23B

This benchmark has not been assessed yet.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

This benchmark has not been assessed yet.

Benchmark task

A plank, computer mouse and blueberry in a steel room test physical stability, spatial relationships, materials and lighting.

Raccoon

Different task instructions

Scene composition and interacting subjects.

The judgment

Beam 1.0 501B A23B

A cheerful, readable illustration with recognizable animals, expressive faces and useful accessories. Simple rounded construction limits complexity, while ambiguous steering attachments and weak hull-to-water contact reduce physical coherence. Lighting and surface effects establish the setting without convincing environmental interaction.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

Expressive animals, a readable riding arrangement and sweeping spray create an appealing cartoon. The raccoon sits convincingly on the dog's back, but the dog's paw does not clearly grip the handlebar, whose connection to the hull is unresolved. Flat lighting provides little volume or reflection. A dark rectangular strip interrupts the upper-right background.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Task & likeness5.0/108.0/10
Anatomy & construction3.5/106.5/10
Space & placement5.5/107.5/10
Light & reflections4.0/105.5/10
Purposeful detail6.5/107.0/10
Artistry & character3.5/107.0/10
Materials & effects5.0/107.5/10
Visual integrity6.5/107.0/10

Benchmark task

A jet-ski scene pairs a canine driver with a raccoon passenger. It tests the placement and contact of interacting subjects, along with vehicle construction.

Pelican

Different task instructions

Subject construction and spatial relationships.

The judgment

Beam 1.0 501B A23B

A cheerful, clearly readable pelican-and-bicycle illustration with clean contours and balanced composition. Simple geometric construction limits artistic complexity, while ambiguous neck and steering overlaps, merged pedal contacts and an incomplete drivetrain weaken physical coherence.

Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)

The pelican is immediately recognizable, with an expressive face, convincing bill silhouette and appealing pastel composition. Both legs connect to feet, and the wing reaches the handlebar. Bicycle construction is less resolved: the far pedal lacks a clear crank connection and no chain is visible. Clean contours and orderly spokes support the illustration, while flat lighting and simplified surfaces limit depth.

Score breakdown

CategoryBeam 1.0 501B A23BPareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)
Task & likeness5.5/108.0/10
Anatomy & construction4.5/106.0/10
Space & placement5.0/107.0/10
Light & reflections4.0/106.0/10
Purposeful detail4.0/106.0/10
Artistry & character4.0/106.0/10
Materials & effects4.0/106.0/10
Visual integrity5.5/108.5/10

Benchmark task

A cycling pelican tests anatomy, riding posture and contact between the bird and the bicycle.