STEALTHMARK · HEAD TO HEAD

All comparisons ↗

DeepSeek-V4.1-Flash vs GPT-6 Astra

GPT-6 Astra leads DeepSeek-V4.1-Flash by 24.7 points on StealthMark, 74.6 to 49.9. It leads on all 9 benchmarks. GPT-6 Astra scores higher on Newton's Mirror, with its largest category advantage in reflections.

See the difference ↘
All results ↗
All results ↗

Overall StealthMark

49.9/100Overall rank 20/58
74.6/100Overall rank 2/58

General Intelligence

48.4/100
73.4/100

Visual Intelligence

53.0/100
76.9/100

Reasoning efficiency

How many reasoning tokens a model needs for the quality it delivers.

Single-step reasoning How much it reasons to understand a task

Multi-step reasoning How much it keeps reasoning as a task goes on

  • GPT-6 Astra

Cost efficiency

How much Performance you get for your money.

Open or local

Open modelMIT

BENCHMARK BY BENCHMARK

Where each model leads

  • DeepSeek-V4.1-Flash
  • GPT-6 Astra
  • Other models
  • Median

DATASHEET

Specs and pricing

DeepSeek-V4.1-Flash is 44× cheaper per output token and runs 3.7× faster in our speed test. GPT-6 Astra scores higher on StealthMark. Both take about 1M tokens of context.

  • DeepSeek-V4.1-Flash
  • GPT-6 Astra
  • Other models
  • Median
PricePer 1M output tokens, input below
DeepSeek-V4.1-Flash: $1.1344× cheaper$0.28 inputGPT-6 Astra: $50$10 input
Context windowTokens per request, max output below
DeepSeek-V4.1-Flash: 1MTied384K max outputGPT-6 Astra: 1.05MTied128K max output
Output speedTokens per second, our measurement
DeepSeek-V4.1-Flash: 2423.7× fastertokens/sGPT-6 Astra: 65tokens/s
SizeTotal and active parameters
DeepSeek-V4.1-Flash: 763B16B activeGPT-6 Astra: Not disclosed
Developer
DeepSeek-AIOpenAI
Released
10 Sep 20263 Sep 2026
Origin
ChinaUnited States
Input → output
Text, image → textBoth
Architecture
20-layer encoder + 20-layer decoder with CSA2, Single-Pass mHC, Engram n-gram memory and DSpark/MTPNot published

Full datasheets, every price and sourceDeepSeek-V4.1-Flash ↗GPT-6 Astra ↗

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.

DeepSeek-V4.1-Flash 55.4/100Rank 15/54

Survived 1 of 3
Foresight: complete playing field
Level 318 ptsTurn 20 / 20

GPT-6 Astra 76.1/100Rank 3/54

Survived 3 of 3
Foresight: complete playing field
Level 318 ptsTurn 20 / 20

Points per level

LevelDeepSeek-V4.1-FlashGPT-6 Astra
Level 1Second Thoughts
16 ptsOut of lives
27 ptsSurvived
Level 2Pressure Pot
7 ptsOut of lives
11 ptsSurvived
Level 3Hide or Flight
18 ptsSurvived
18 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.

DeepSeek-V4.1-Flash 65.2/100Rank 24/55

Cleared 5 of 5
Latch: complete playing field
Level 520.6 ptsTurn 27 / 40

GPT-6 Astra 87.6/100Rank 2/55

Cleared 5 of 5
Latch: complete playing field
Level 520.6 ptsTurn 27 / 40

Points per level

LevelDeepSeek-V4.1-FlashGPT-6 Astra
Level 1Hold the door
5.2 ptsCleared
5.2 ptsCleared
Level 2Crossfeed
9 ptsCleared
10.2 ptsClearedFinished early · 1 beacon left
Level 3Relay vault
17.2 ptsClearedFinished early · 3 beacons left
19.2 ptsCleared
Level 4Passing bay
18.2 ptsClearedFinished early · 3 beacons left
18.2 ptsClearedFinished early · 3 beacons left
Level 5Second thoughts
20.6 ptsCleared
20.6 ptsCleared

Newton's Mirror

Same task instructions

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

The judgment

DeepSeek-V4.1-Flash

The first wall and floor images provide limited reflection geometry, but the room's mirror recurrence is substantially incomplete, particularly the outer-edge chains and lateral sightlines. The main sphere and its large reflected images lack ordered, curved room and virtual-sphere imagery. Repeated emitter mapping is missing from both planar mirrors and spherical surfaces.

GPT-6 Astra

Major placement and the visible contact sequence are convincing. Dense planar light repetition, recognizable spherical panel images and simplified nested sphere motifs provide useful optical structure. Missing outer chains, truncated distant sequences and incorrect bead-like belts still leave the recursive reflections substantially incomplete.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
State at 600 ms23.5/2525.0/25
Camera and placement12.5/1515.0/15
Reflections30.0/12070.0/120
Materials and scale8.0/1010.0/10
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.

The judgment

DeepSeek-V4.1-Flash

Expressive faces, cobalt coloring and a flowing diagonal composition make the mutual inspection charming. Execution is uneven: the long pointed flipper lacks a convincing grip, the enlarged shrimp face appears detached from its body, and the large lens is oval. Flat anatomical shading and an unexplained purple ringed patch beneath the whale weaken the requested polished depth.

GPT-6 Astra

The sixfold silhouette, pale coloring, outward heads and central hexagon closely match the brief. Fine scales and soft shading support rounded volume. However, several crossings have conspicuous rectangular seams and weak contact shadows, making the weave feel assembled. Small heads have limited anatomical definition, and the repetitive texture conveys less organic life than requested.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
Task & likeness7.0/109.0/10
Anatomy & construction6.5/108.0/10
Space & placement6.5/107.5/10
Light & reflections6.0/107.0/10
Purposeful detail7.0/108.0/10
Artistry & character7.5/108.5/10
Materials & effects7.0/108.0/10
Visual integrity7.0/108.5/10

Benchmark task

DeepSeek-V4.1-Flash

A whale and a tiny shrimp examine one another through bubble lenses, with an octopus nearby. The scene tests interacting characters, differences in scale and transparent materials.

GPT-6 Astra

A pale serpent woven into an interlocking knot tests connected anatomy, repeated structure and readable overlaps. Organic surface detail must work with the overall geometry.

Mona Lisa

Different task instructions

Precise visual recall and portrait reconstruction.

The judgment

DeepSeek-V4.1-Flash

The centered portrait, dark hair, crossed hands and muted landscape make the reference recognizable. Replication is limited by the frontal pose, staring eyes and weakly defined smile. Hands are overlapping oval masses with almost no finger articulation, while wrists disappear into sleeves. Soft gradients suggest volume, but broad, indistinct surfaces and faint stray lines leave the drawing basic and uneven.

GPT-6 Astra

Controlled hair contours, neckline detail, modeled sleeves and a layered river landscape give this recognizable portrait developed illustrative finish. The elongated face, largely frontal torso, crowded fingers and weakly established chair support limit replication, anatomical precision and spatial conviction. Small construction slips at the neck and jaw, a stray dark shape under the right cuff and an abruptly ending armrest bar weaken integrity.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
Task & likeness3.5/106.5/10
Anatomy & construction3.0/106.5/10
Space & placement4.5/106.5/10
Light & reflections4.5/106.5/10
Purposeful detail3.0/106.5/10
Artistry & character4.0/107.0/10
Materials & effects3.0/107.0/10
Visual integrity5.5/107.5/10

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

DeepSeek-V4.1-Flash

The glowing sunset, silhouetted palm and playful foreground create a coherent, appealing beach illustration. Birds, crab, boat and reflections fulfill the brief. The airborne animals have weakly defined anatomy and little convincing splash interaction. Broad, blurred reflections and largely flat surfaces limit depth and finish.

GPT-6 Astra

A richly composed sunset beach illustration with cohesive peach, mauve and teal colors. Sweeping palm fronds, a scalloped umbrella and curved shorelines frame the sun and its aligned reflection. Distinct animal silhouettes and carefully distributed beach details reward inspection. Simplified human and dog anatomy, slight dolphin splash cues and inconsistent equipment shadows limit physical conviction.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
Task & likeness8.0/109.0/10
Anatomy & construction5.5/107.5/10
Space & placement6.5/108.5/10
Light & reflections6.5/107.5/10
Purposeful detail5.5/108.5/10
Artistry & character5.0/108.5/10
Materials & effects6.0/107.5/10
Visual integrity5.5/107.5/10

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

DeepSeek-V4.1-Flash

This benchmark has not been assessed yet.

GPT-6 Astra

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

DeepSeek-V4.1-Flash

The sunset palette and silhouetted grouping establish an appealing scene. A masked rider sits above a pointed-eared dog, but merged body shapes leave limbs and support unresolved. The boatlike craft lacks readable steering controls, so the requested action is unclear. Sparse facial marks and glowing spray dots offer limited detail.

GPT-6 Astra

A controlled, expressive illustration with readable riding and steering, articulated poses and purposeful small details. Layered foam and coherent hull perspective establish motion and depth. Simplified paws, compressed control fittings and limited directional lighting constrain its physical sophistication.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
Task & likeness5.0/108.5/10
Anatomy & construction4.0/107.5/10
Space & placement5.5/107.5/10
Light & reflections6.0/106.5/10
Purposeful detail3.5/108.0/10
Artistry & character4.0/107.5/10
Materials & effects5.0/107.5/10
Visual integrity6.0/108.0/10

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

DeepSeek-V4.1-Flash

A recognizable pelican rides a brightly colored bicycle against a striking sunset. The sweeping bill, trailing scarf and circular wheels create a controlled composition. Both legs connect, but pedal linkage, body support and the wing's handlebar grip remain loosely resolved. Clean vector shapes and feather details work well; simplified shading limits volume and integration with the sunset light.

GPT-6 Astra

The pelican's long bill, flowing scarf and warm coastal palette create a charming, balanced illustration. Wheels, spokes and frame are largely coherent, but neither wing reaches the handlebars, and the rear leg's pedal contact is unclear. Flat lighting and simplified surfaces limit depth despite clean outlines and attractive feather detail.

Score breakdown

CategoryDeepSeek-V4.1-FlashGPT-6 Astra
Task & likeness8.5/108.0/10
Anatomy & construction7.0/107.0/10
Space & placement7.5/107.5/10
Light & reflections7.0/107.0/10
Purposeful detail6.5/108.0/10
Artistry & character8.0/108.0/10
Materials & effects7.0/107.5/10
Visual integrity8.0/108.0/10

Benchmark task

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