STEALTHMARK · HEAD TO HEAD

All comparisons ↗

Qwen3.5-9B vs Step 5 Preview

Step 5 Preview leads Qwen3.5-9B by 28.1 points on StealthMark, 39.0 to 10.9. It leads on all 9 benchmarks. Step 5 Preview scores higher on Newton's Mirror, with its largest category advantage in reflections.

See the difference ↘
All results ↗
All results ↗

Overall StealthMark

10.9/100Overall rank 53/53
39.0/100Overall rank 36/53

General Intelligence

7.9/100
37.1/100

Visual Intelligence

15.6/100
41.9/100

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

Open or local

Local modelOpen model · Apache-2.0

BENCHMARK BY BENCHMARK

Where each model leads

  • Qwen3.5-9B
  • Step 5 Preview
  • Other models
  • Median

DATASHEET

Specs and pricing

Both take about 1.01M tokens of context.

  • Qwen3.5-9B
  • Step 5 Preview
  • Other models
  • Median
PricePer 1M output tokens, input below
Qwen3.5-9B: $0.25$0.17 inputStep 5 Preview: No list price
Context windowTokens per request, max output below
Qwen3.5-9B: 1.01MTiedStep 5 Preview: 1MTied
Output speedTokens per second, our measurement
Qwen3.5-9B: 161tokens/sStep 5 Preview: Not measured
SizeTotal and active parameters
Qwen3.5-9B: 9BAll activeEstimatedStep 5 Preview: 600B27B active
Developer
QwenStepFun
Released
2 Mar 2026Not published
Origin
ChinaNot published
Input → output
Text, image, video → textBoth
Architecture
Dense causal language model with vision encoder; Gated DeltaNet + gated attention; MTP-trainedSparse mixture-of-experts; 92 layers (88 MoE), 352 routed experts with 8 selected per token plus a shared expert; grouped-query attention with 512-token sliding windows and full attention every fourth layer; 3 multi-token-prediction layers; vision tower for image and video input

Full datasheets, every price and sourceQwen3.5-9B ↗Step 5 Preview ↗

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.

Qwen3.5-9B 12.8/100Rank 45/49

Survived 0 of 3
Foresight: complete playing field
Level 32 ptsTurn 8 / 20

Step 5 Preview 59.9/100Rank 42/491 of 3 levels

1 of 3 played
Foresight: complete playing field
Level 122 ptsTurn 28 / 28

Points per level

LevelQwen3.5-9BStep 5 Preview
Level 1Second Thoughts
4 ptsOut of lives
22 ptsSurvived
Level 2Pressure Pot
4 ptsOut of lives
Pending
Level 3Hide or Flight
2 ptsOut of lives
Pending

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.

Qwen3.5-9B 3.8/100Rank 45/50

Cleared 1 of 5
Latch: complete playing field
Level 54.4 ptsTurn 33 / 40

Step 5 Preview 40.7/100Rank 28/50

Cleared 4 of 5
Latch: complete playing field
Level 515 ptsTurn 35 / 40

Points per level

LevelQwen3.5-9BStep 5 Preview
Level 1Hold the door
5 ptsCleared
5.2 ptsCleared
Level 2Crossfeed
3.8 ptsFinished early
9 ptsCleared
Level 3Relay vault
1.2 ptsFinished early
18.6 ptsCleared
Level 4Passing bay
0.2 ptsFinished early
15.6 ptsClearedFinished early · 3 beacons left
Level 5Second thoughts
4.4 ptsFinished early
15 ptsFinished early

Newton's Mirror

Same task instructions

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

The judgment

Qwen3.5-9B

Horizontal centering and a separate airborne black blob are visible. Cyan's recent contact and crimson's forward transport are insufficiently demonstrated, while sphere support remains obscured. The enlarged, lowered sphere substantially changes the composition. The wall and floor sphere images, recursive room structure and curved emitter mappings are almost entirely missing.

Step 5 Preview

The main sphere, airborne black and small colored features occupy broadly appropriate positions. Most of the defining mirror system is absent, and paint interaction remains only partially supported. The small paint features do not visibly establish a separate volume-scale violation.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
State at 600 ms3.0/2520.0/25
Camera and placement4.5/1513.5/15
Reflections2.0/12020.0/120
Materials and scale3.0/107.0/10
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.

The judgment

Qwen3.5-9B

Glowing teal spheres and a harmonious palette provide visible strengths, but the central character remains seriously unresolved. A large gap separates head and torso, limbs cannot be traced, and scattered garment shapes fail to establish the ornate costume. The foreground mound replaces the tall enclosing arch. Clean primitive edges do not repair these construction failures.

Step 5 Preview

A charming diagonal climber and warm palette communicate the central idea. Simple rounded construction, crowded limb counting, uncertain footholds and an unattached upper shelf limit its physical coherence. Growth bands and the yarn trail add specificity, but the requested material distinctions and unravelled bootie remain underdeveloped.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
Task & likeness3.0/106.0/10
Anatomy & construction2.0/104.5/10
Space & placement3.0/105.0/10
Light & reflections3.5/105.5/10
Purposeful detail2.0/105.5/10
Artistry & character2.0/105.0/10
Materials & effects4.0/106.0/10
Visual integrity3.0/105.0/10

Benchmark task

Qwen3.5-9B

A floating capybara in ornate robes tests character anatomy, layered fabric and the placement of surrounding luminous forms.

Step 5 Preview

Character scenes test expression, anatomy, materials and coherent visual construction.

Mona Lisa

Different task instructions

Precise visual recall and portrait reconstruction.

The judgment

Qwen3.5-9B

A centered dark oval with sparse facial marks forms a simple graphic face, but scarcely replicates the Mona Lisa. The portrait's body, crossed hands, clothing and landscape are missing, while angular facial construction and flat tones lose its likeness, volume and subtle expression.

Step 5 Preview

A recognizable, coherent simplification with purposeful facial contours and differentiated fingers. These provide more construction and detail than the broad silhouette alone suggests. The oversized face, frontal seated pose, unresolved wrists and sparse landscape still substantially limit likeness, depth and finish.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
Task & likeness0.0/104.0/10
Anatomy & construction0.5/104.0/10
Space & placement0.0/104.0/10
Light & reflections0.0/103.5/10
Purposeful detail0.5/104.0/10
Artistry & character1.0/104.0/10
Materials & effects0.0/103.0/10
Visual integrity0.5/103.5/10

Benchmark task

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

Beach

Same task instructions

Creativity, perspective and visual beauty.

The judgment

Qwen3.5-9B

The beach is recognizable, and sweeping shoreline curves provide basic rhythm. Wildlife and playful activity remain poorly resolved, with ambiguous body construction and object contact. The sun glows without corresponding reflections or surface lighting. Fragmented subjects and an unintegrated rectangular panel substantially weaken composition and finish.

Step 5 Preview

A recognizable sunset beach with clear warm-cool contrast and playful accessories. Basic animal construction, shallow foreground placement and schematic reflections limit spatial and physical credibility. Sparse descriptive detail and abruptly ended sky bands constrain the finish.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
Task & likeness1.5/104.0/10
Anatomy & construction0.5/103.0/10
Space & placement1.5/103.5/10
Light & reflections1.0/103.5/10
Purposeful detail2.5/103.0/10
Artistry & character1.5/103.0/10
Materials & effects0.5/103.5/10
Visual integrity1.5/104.0/10

Benchmark task

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

Ground Truth

Same task instructions

Physical reasoning, light and visual reconstruction.

The judgment

Qwen3.5-9B

This benchmark has not been assessed yet.

Step 5 Preview

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

Same task instructions

Scene composition and interacting subjects.

The judgment

Qwen3.5-9B

Bright colors, clean individual shapes and basic animal cues make the scene partly recognizable. Back seating, body continuity, steering and jet ski construction remain unresolved. Background gradients and wave rhythm provide some finish, but substantial anatomical, spatial and interaction weaknesses dominate.

Step 5 Preview

Recognizable animals, a bright palette and identifying accessories support a basic cartoon. Primitive anatomy, partial seated support, unresolved steering and weak hull-water contact limit the central action. Lighting and spray provide atmosphere without establishing convincing physical depth.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
Task & likeness2.5/104.0/10
Anatomy & construction1.0/104.5/10
Space & placement2.0/103.0/10
Light & reflections1.0/103.5/10
Purposeful detail1.5/104.0/10
Artistry & character1.0/103.5/10
Materials & effects1.0/103.0/10
Visual integrity2.5/103.0/10

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

Same task instructions

Subject construction and spatial relationships.

The judgment

Qwen3.5-9B

The oversized bill and orderly spoked wheels make the subject partly recognizable. Clean contours and strong color contrast provide graphic clarity, but unresolved appendages, steering, drivetrain and support dominate the construction. Sparse shading and detail further limit the illustration.

Step 5 Preview

The pelican and bicycle are recognizable, with a pleasant palette and basic diffuse shading. The riding relationship fails physically: the bird is unsupported, the front wheel is unmounted, and pedalling is unresolved. Simple geometric construction and sparse material treatment further limit the drawing.

Score breakdown

CategoryQwen3.5-9BStep 5 Preview
Task & likeness4.0/103.0/10
Anatomy & construction1.5/103.0/10
Space & placement3.0/103.0/10
Light & reflections1.5/103.0/10
Purposeful detail2.5/104.0/10
Artistry & character1.5/103.0/10
Materials & effects1.5/103.0/10
Visual integrity2.5/102.0/10

Benchmark task

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