Which Ollama Local Model is Best? Nautilus Shell SVG Comparison (15 Models)

The nautilus shell is the iconic embodiment of the golden spiral – a logarithmic spiral whose growth factor is phi (the golden ratio, approximately 1.618). We asked our models to draw a nautilus shell to test whether they can render this most mathematical of natural forms. A real nautilus shell is a perfect logarithmic spiral with chambers, a pearlescent surface with brown zebra stripes, and a hollow center where the spiral begins. This prompt tests curve precision (the spiral must expand smoothly), color gradients (pearlescent white to brown stripes), and structural detail (chambers inside the spiral).

The prompt was: Make an svg image of a nautilus shell

This is the fourteenth in our SVG benchmark series. See also: duck on a bicycle, duck with a parachute, duck driving a jeep, cherry blossom trees, duck programmer debugging at 3am, baby shark fish, octopus playing chess, FIFA World Cup 2026, elephant on a skateboard, flying helicopter, sunflower with seeds, pineapple, pinecone, dates palm tree.

Why a nautilus shell? The nautilus shell prompt is a curve precision stress test because it combines: (1) Logarithmic spiral – the shell’s shape is a perfect golden spiral that grows by phi each quarter turn; a model with mathematical intuition should use <path> with Bezier curves to approximate this smooth expansion, (2) Chamber divisions – a real nautilus shell has internal chambers separated by walls; the model should show these as curved lines crossing the spiral, (3) Pearlescent coloring – the shell surface has iridescent white/cream tones with brown zebra-like stripes, (4) Hollow center – the spiral opens from a tiny center to a large opening, (5) Bilateral symmetry of stripes – the brown stripes on the exterior follow the curve of the shell, and (6) Curve smoothness – the spiral must be smooth, not angular or polygonal.

The goal is not to declare a winner – it is to give you the data so you can pick the best model for your own use case. We show you the SVG, the stats, and a short analysis for each. You decide.

How to Choose the Best Ollama Model for Nautilus Shell SVGs

The nautilus shell prompt rewards different things than previous prompts. Here are the criteria to use:

• Spiral accuracy: Does the model draw a smooth logarithmic spiral that expands gradually? Or is it a flat circle / angular polygon?
• Chamber divisions: Does it show internal chamber walls as curved lines crossing the spiral?
• Stripe pattern: Does it render brown zebra stripes on the exterior following the curve?
• Color gradient: Does it use pearlescent white/cream tones with shading? Or flat color?
• Spiral smoothness: Is the spiral drawn with smooth Bezier curves? Or straight line segments?
• Hollow center: Does the spiral start from a tiny center point and expand outward?
• SVG code quality: Does it use <path> with d attribute curves (C, Q, A commands)? Or just straight lines?

How It Works

The script discovers all locally installed models via the Ollama API (/api/tags), then sends the identical prompt through the OpenAI-compatible endpoint (http://localhost:11434/v1/chat/completions). Each model’s response is parsed for an <svg>...</svg> block, and the extracted SVG is saved for rendering with minimal post-processing (adding width="100%" height="auto" for responsive embedding and fixing XML errors so the SVG renders in browsers).

Unlike our cloud model benchmarks, these models run entirely on the local GPU – no cloud subscription or network round-trip required. This means generation times reflect local hardware performance, and model sizes range from 1B to 31B parameters. Embedding, vision, and OCR models are automatically skipped.

Summary Table: Compare All Models at a Glance

Use this table to quickly compare models on the metrics that matter. The verdict column is a one-line summary to help you shortlist – but read the per-model sections below for the full picture before you decide.

#	Model	SVG Size	Shapes	Colors	Complexity	Verdict
1	cieloforge/qwen2.5-14B-instruct-spec:latest	1355	5	2	Compact	Compact
2	deepseek-r1:1.5b	1849	4	0	Compact	Compact
3	deepseek-r1:7b	207	0	0	Compact	Compact
4	gemma3:1b-it-qat	544	3	4	Compact	Compact
5	gemma3:4b	685	3	3	Compact	Compact
6	gemma4:12b	1131	6	4	Compact	Compact
7	gemma4:26b-a4b-it-qat	2258	7	9	Compact	Compact
8	lfm2.5:latest	1074	5	10	Compact	Compact
9	llama3.1:8b	675	2	0	Compact	Compact
10	qwen2.5-coder-14b:latest	708	2	3	Compact	Compact
11	qwen2.5-coder:7b	568	2	0	Compact	Compact
12	qwen2.5:7b	331	1	2	Compact	Compact
13	qwen3.5:4b	1875	5	5	Compact	Compact
14	qwen3.5:9b	1867	6	7	Compact	Compact
15	qwen3.5:latest	1047	2	4	Compact	Compact
16	gemma4:31b-it-qat	-	-	-	-	Failed
17	gemma4:latest	-	-	-	-	Failed
18	jeffgreen311/Eve-V2-Unleashed-Qwen3.5-8B-Liberated-4K-4B-Merged:latest	-	-	-	-	Failed
19	kwangsuklee/Qwen3.5-9B.Q4_K_M-Claude-4.6-Opus-Reasoning-Distilled-v2:latest	-	-	-	-	Failed
20	qwen2.5:3b	-	-	-	-	Failed
21	qwen3:14b	-	-	-	-	Failed
22	qwen3:8b	-	-	-	-	Failed
23	SetneufPT/Qwen3.6-27B-MTP_Q3_32K_16GB-GPU:latest	-	-	-	-	Failed
24	SetneufPT/Qwen3.6-27B.MTP_Q3_32K_16GB-GPU:latest	-	-	-	-	Failed
25	VladimirGav/Qwen3.6-27B-16GB-VRAM-Uncensored:latest	-	-	-	-	Failed

15 out of 25 models produced a valid SVG. The 10 that failed either returned an error or did not include a valid <svg>...</svg> block in their response.

Quick Recommendation by Use Case

If you just want a shortcut, here is which model to pick based on what you care about:

• You want the most mathematically accurate golden spiral: look for models whose SVG shows a smooth expanding spiral curve
• You want chamber divisions for biological accuracy: check the per-model analysis for internal chamber walls
• You want the most visually rich shell: pick models labeled “Very high” complexity in the table above
• You want smooth Bezier curve rendering: look for models that used <path> with curve commands
• You want a small SVG for web embedding: pick models with “Compact” verdict
• You want a balance of detail and speed: pick models labeled “Balanced” or “Detailed”

Now read on for the full per-model breakdown and judge for yourself.

1. cieloforge/qwen2.5-14B-instruct-spec:latest

SVG size: 1355 characters
Complexity: Compact
Shape elements: 5
Distinct colors: 2
Raw response: 2121 characters
Generation time: 142.8s

Analysis

This SVG contains approximately 5 shape elements and uses 2 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

2. deepseek-r1:1.5b

SVG size: 1849 characters
Complexity: Compact
Shape elements: 4
Distinct colors: 0
Raw response: 2708 characters
Generation time: 146.0s

Analysis

This SVG contains approximately 4 shape elements and uses 0 distinct colors.

Notable SVG techniques used:

• Uses transforms (translate, rotate, scale) for precise positioning and rotation
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

4. gemma3:1b-it-qat

SVG size: 544 characters
Complexity: Compact
Shape elements: 3
Distinct colors: 4
Raw response: 3147 characters
Generation time: 142.9s

Analysis

This SVG contains approximately 3 shape elements and uses 4 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

5. gemma3:4b

SVG size: 685 characters
Complexity: Compact
Shape elements: 3
Distinct colors: 3
Raw response: 3892 characters
Generation time: 176.3s

Analysis

This SVG contains approximately 3 shape elements and uses 3 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

6. gemma4:12b

SVG size: 1131 characters
Complexity: Compact
Shape elements: 6
Distinct colors: 4
Raw response: 2393 characters
Generation time: 134.6s

Analysis

This SVG contains approximately 6 shape elements and uses 4 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

7. gemma4:26b-a4b-it-qat

SVG size: 2258 characters
Complexity: Compact
Shape elements: 7
Distinct colors: 9
Raw response: 3331 characters
Generation time: 284.2s

Analysis

This SVG contains approximately 7 shape elements and uses 9 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth
• Uses transforms (translate, rotate, scale) for precise positioning and rotation
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

8. lfm2.5:latest

SVG size: 1074 characters
Complexity: Compact
Shape elements: 5
Distinct colors: 10
Raw response: 2165 characters
Generation time: 424.8s

Analysis

This SVG contains approximately 5 shape elements and uses 10 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

9. llama3.1:8b

SVG size: 675 characters
Complexity: Compact
Shape elements: 2
Distinct colors: 0
Raw response: 2336 characters
Generation time: 202.2s

Analysis

This SVG contains approximately 2 shape elements and uses 0 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

10. qwen2.5-coder-14b:latest

SVG size: 708 characters
Complexity: Compact
Shape elements: 2
Distinct colors: 3
Raw response: 1474 characters
Generation time: 241.3s

Analysis

This SVG contains approximately 2 shape elements and uses 3 distinct colors.

Notable SVG techniques used:

• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

11. qwen2.5-coder:7b

SVG size: 568 characters
Complexity: Compact
Shape elements: 2
Distinct colors: 0
Raw response: 1701 characters
Generation time: 230.9s

Analysis

This SVG contains approximately 2 shape elements and uses 0 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

12. qwen2.5:7b

SVG size: 331 characters
Complexity: Compact
Shape elements: 1
Distinct colors: 2
Raw response: 1552 characters
Generation time: 169.6s

Analysis

This SVG contains approximately 1 shape elements and uses 2 distinct colors.

Notable SVG techniques used:

• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

13. qwen3.5:4b

SVG size: 1875 characters
Complexity: Compact
Shape elements: 5
Distinct colors: 5
Raw response: 2169 characters
Generation time: 143.0s

Analysis

This SVG contains approximately 5 shape elements and uses 5 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Uses <pattern> elements for repeating textures (e.g., pineapple skin or trunk scars)
• Includes gradient fills for richer visual depth
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

14. qwen3.5:9b

SVG size: 1867 characters
Complexity: Compact
Shape elements: 6
Distinct colors: 7
Raw response: 1863 characters
Generation time: 156.3s

Analysis

This SVG contains approximately 6 shape elements and uses 7 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth
• Uses transforms (translate, rotate, scale) for precise positioning and rotation
• Uses Bezier curves (C, Q, A path commands) for smooth organic shapes
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

15. qwen3.5:latest

SVG size: 1047 characters
Complexity: Compact
Shape elements: 2
Distinct colors: 4
Raw response: 3793 characters
Generation time: 440.7s

Analysis

This SVG contains approximately 2 shape elements and uses 4 distinct colors.

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth
• Contains XML comments, showing the model’s internal documentation of its work

This is a compact SVG (under 5,000 characters) that uses minimal markup to convey the scene. The model prioritized efficiency and simplicity over elaborate detail.

View raw SVG source

```xml ```

Failed Models

The following 10 models did not produce a valid SVG (either returned an error or the response did not contain a parseable <svg>...</svg> block):

• gemma4:31b-it-qat – TimeoutError: timed out
• gemma4:latest – TimeoutError: timed out
• jeffgreen311/Eve-V2-Unleashed-Qwen3.5-8B-Liberated-4K-4B-Merged:latest – TimeoutError: timed out
• kwangsuklee/Qwen3.5-9B.Q4_K_M-Claude-4.6-Opus-Reasoning-Distilled-v2:latest – no SVG block in response
• qwen2.5:3b – no SVG block in response
• qwen3:14b – TimeoutError: timed out
• qwen3:8b – TimeoutError: timed out
• SetneufPT/Qwen3.6-27B-MTP_Q3_32K_16GB-GPU:latest – TimeoutError: timed out
• SetneufPT/Qwen3.6-27B.MTP_Q3_32K_16GB-GPU:latest – TimeoutError: timed out
• VladimirGav/Qwen3.6-27B-16GB-VRAM-Uncensored:latest – TimeoutError: timed out

Conclusion

We asked 25 Ollama local models to draw a nautilus shell – a subject deeply connected to Fibonacci mathematics. The results reveal each model’s natural instinct for mathematical patterns, organic curves, and natural color palettes.

Key takeaways:

• Mathematical intuition varies widely: some models attempted Fibonacci spiral arrangements naturally, while others defaulted to simple grids or flat shapes. The models that attempted spirals demonstrate a deeper understanding of natural geometry.
• Detail and file size trade-off: models that produced the richest scenes (Very high complexity) also generated the largest SVG files. For web embedding, “Balanced” or “Detailed” models may be more practical.
• Color palettes differ dramatically: some models used sophisticated gradients and 15+ distinct colors, while others used as few as 4 flat colors. More colors generally means a more lifelike result.
• Code structure quality varies: the best models used <defs>, <use>, and <pattern> elements to efficiently generate repeating structures (seeds, scales, leaflets). This is a strong signal of model capability for practical SVG work.
• No single model is best at everything: the “right” model depends on whether you prioritize mathematical accuracy, visual beauty, file size, or code quality.

Try it yourself: run the benchmark script with your own prompt and see how the models perform on your specific use case. The full code is available in the ollama-svg-benchmark repository.

---

Which model do you think drew the best nautilus shell? Let us know in the comments below!

Enjoyed this post? Never miss out on future posts by following us