Which Ollama Cloud Model is Best? Baby Shark Fish SVG Comparison (13 Models)

After testing LLMs on ducks, vehicles, dev scenes, and nature, we wanted to know: can today’s top models draw marine life? This time we asked 13 Ollama cloud models to draw a baby shark fish – a prompt that tests organic curves, underwater atmosphere, character anatomy, and cute styling.

The prompt was: Make an svg image of a baby shark fish

This is the sixth in our SVG benchmark series. See also: duck on a bicycle, duck with a parachute, duck driving a jeep, cherry blossom trees, and duck programmer debugging at 3am.

Why a baby shark? Unlike previous prompts, this scene combines multiple hard problems for SVG generation: (1) organic curved shapes (shark body, fins, tail) – harder than rectangles, (2) underwater atmosphere (bubbles, gradients, waves), (3) anatomical correctness (dorsal fin, pectoral fins, gills, caudal tail), (4) character emotion (the “baby” cue implies cute, rounded proportions with big eyes), and (5) scene context (coral, seaweed, small fish). A model that draws a great jeep may struggle with organic curves, and vice versa – which is exactly why we run multiple prompts.

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 Animal/Marine SVGs

Marine-life prompts reward different things than indoor or vehicle prompts. Here are the criteria to use:

• Anatomical accuracy: Does the SVG have a shark body, dorsal fin, pectoral fins, tail, and gills? Or is it a generic fish shape?
• Organic curves: Sharks have smooth, curved bodies. Does the model use <path> with bezier curves, or only straight lines and rectangles?
• Underwater atmosphere: Look for bubbles, waves, gradients, or blue color palettes that convey water. A shark on a white background misses the scene.
• Baby/cute styling: The prompt says “baby.” Does the model render a cute, rounded shark with big eyes, or a scary adult shark? Prompt adherence matters.
• Scene context: Coral, seaweed, small fish, or a reef add depth and show the model understood the habitat.
• SVG code quality: Does it use <defs>, <use>, gradients, and clean structure? Better code is easier to tweak (e.g., to recolor or animate).

How It Works

The script discovers all cloud-hosted models via the Ollama API (/api/tags), pulls each model, 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 zero post-processing (other than adding width="100%" height="auto" for responsive embedding and fixing any XML errors so the SVG renders in browsers).

Cloud models are identified by the remote_host field in the API response – these models are hosted on Ollama Cloud rather than running locally. This means even very large models (671B parameters) can be queried instantly without local GPU resources.

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	deepseek-v4-flash:cloud	4917	42	15	High	Detailed
2	deepseek-v4-pro:cloud	3287	27	8	Medium	Balanced
3	gemma4:31b-cloud	1627	15	9	Medium	Balanced
4	gemma4:cloud	1615	13	8	Compact	Compact
5	glm-5.1:cloud	11903	69	23	Very high	Richest scene
6	glm-5.2:cloud	6479	43	14	High	Detailed
7	gpt-oss:120b-cloud	1212	8	6	Compact	Compact
8	kimi-k2.6:cloud	776	3	0	Compact	Compact
9	minimax-m2.7:cloud	2341	18	10	Medium	Balanced
10	minimax-m3:cloud	2859	24	6	Medium	Balanced
11	nemotron-3-super:cloud	938	7	5	Compact	Compact
12	nemotron-3-ultra:cloud	5199	25	8	Medium	Detailed
13	qwen3.5:397b-cloud	2023	19	7	Medium	Balanced
14	bjoernb/claude-opus-4-5:latest	-	-	-	-	Retired (410)
15	deepseek-v3.1:671b-cloud	-	-	-	-	Retired (410)
16	glm-5:cloud	-	-	-	-	Retired (410)
17	qwen3-vl:235b-cloud	-	-	-	-	Retired (410)

13 out of 17 active models produced a valid SVG. The 4 retired models returned HTTP 410 Gone (removed from Ollama Cloud on 2026-07-15).

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 detailed shark scene SVG: pick models labeled “Very high” complexity in the table above
• You want the fastest response: look at the per-model sections below for the elapsed time
• You want the cleanest, most reusable SVG code: pick models that use <defs>, <use>, and transforms (see raw source below each SVG)
• You want a small, efficient SVG for web embedding: pick models with “Compact” verdict
• You want a balance of detail and speed: pick models labeled “Balanced” or “Detailed”
• You want to compare within a model family: pick deepseek-v4-pro vs deepseek-v4-flash, or glm-5.1 vs glm-5.2, or minimax-m2.7 vs minimax-m3

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

1. deepseek-v4-flash:cloud

SVG size: 4917 characters
Complexity: High
Shape elements: 42
Distinct colors: 15
Raw response: 5311 characters
Generation time: 72.6s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Applies SVG filters (e.g., drop shadows, blurs) for atmospheric effects
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat
• Coral/seaweed is present – the model decorated the underwater scene
• Baby/cute styling is present – the model captured the “baby” cue

With 4,917 characters of SVG markup, this is a balanced output – enough detail to be recognizable without being bloated.

View raw SVG source

```xml ```

2. deepseek-v4-pro:cloud

SVG size: 3287 characters
Complexity: Medium
Shape elements: 27
Distinct colors: 8
Raw response: 3656 characters
Generation time: 59.3s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth (great for ocean water and lighting)
• Applies SVG filters (e.g., drop shadows, blurs) for atmospheric effects
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat
• Baby/cute styling is present – the model captured the “baby” cue

With 3,287 characters of SVG markup, this is a balanced output – enough detail to be recognizable without being bloated.

View raw SVG source

```xml ```

3. gemma4:31b-cloud

SVG size: 1627 characters
Complexity: Medium
Shape elements: 15
Distinct colors: 9
Raw response: 2322 characters
Generation time: 8.6s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Bubbles are present – the model created an underwater atmosphere
• Baby/cute styling is present – the model captured the “baby” cue

With 1,627 characters of SVG markup, this is a compact output – minimal but potentially efficient for web embedding.

View raw SVG source

```xml ```

4. gemma4:cloud

SVG size: 1615 characters
Complexity: Compact
Shape elements: 13
Distinct colors: 8
Raw response: 2428 characters
Generation time: 44.5s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Baby/cute styling is present – the model captured the “baby” cue

With 1,615 characters of SVG markup, this is a compact output – minimal but potentially efficient for web embedding.

View raw SVG source

```xml ```

5. glm-5.1:cloud

SVG size: 11903 characters
Complexity: Very high
Shape elements: 69
Distinct colors: 23
Raw response: 16204 characters
Generation time: 12.0s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth (great for ocean water and lighting)
• Applies SVG filters (e.g., drop shadows, blurs) for atmospheric effects
• Uses transforms (translate, rotate, scale) for positioning the shark and scene elements
• Includes animation elements – the model attempted motion (e.g., swimming, bubbles)
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat
• Small fish are present – the model added scene context
• Coral/seaweed is present – the model decorated the underwater scene
• Baby/cute styling is present – the model captured the “baby” cue

With 11,903 characters of SVG markup, this is a detailed output that balances richness with readability.

View raw SVG source

```xml ```

6. glm-5.2:cloud

SVG size: 6479 characters
Complexity: High
Shape elements: 43
Distinct colors: 14
Raw response: 6526 characters
Generation time: 63.8s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth (great for ocean water and lighting)
• Applies SVG filters (e.g., drop shadows, blurs) for atmospheric effects
• Uses transforms (translate, rotate, scale) for positioning the shark and scene elements
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat
• Small fish are present – the model added scene context
• Coral/seaweed is present – the model decorated the underwater scene
• Baby/cute styling is present – the model captured the “baby” cue

With 6,479 characters of SVG markup, this is a detailed output that balances richness with readability.

View raw SVG source

```xml ```

7. gpt-oss:120b-cloud

SVG size: 1212 characters
Complexity: Compact
Shape elements: 8
Distinct colors: 6
Raw response: 1934 characters
Generation time: 6.9s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character

With 1,212 characters of SVG markup, this is a compact output – minimal but potentially efficient for web embedding.

View raw SVG source

```xml ```

8. kimi-k2.6:cloud

SVG size: 776 characters
Complexity: Compact
Shape elements: 3
Distinct colors: 0
Raw response: 17103 characters
Generation time: 51.4s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth (great for ocean water and lighting)

Concept adherence: Limited – the model may have drawn a generic shape rather than a recognizable baby shark.

With 776 characters of SVG markup, this is a compact output – minimal but potentially efficient for web embedding.

View raw SVG source

```xml ```

9. minimax-m2.7:cloud

SVG size: 2341 characters
Complexity: Medium
Shape elements: 18
Distinct colors: 10
Raw response: 2520 characters
Generation time: 59.3s

Analysis

Notable SVG techniques used:

• Uses transforms (translate, rotate, scale) for positioning the shark and scene elements
• Includes animation elements – the model attempted motion (e.g., swimming, bubbles)
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat

With 2,341 characters of SVG markup, this is a balanced output – enough detail to be recognizable without being bloated.

View raw SVG source

```xml ```

10. minimax-m3:cloud

SVG size: 2859 characters
Complexity: Medium
Shape elements: 24
Distinct colors: 6
Raw response: 3450 characters
Generation time: 13.6s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Water/ocean scene is present – the model understood the habitat
• Baby/cute styling is present – the model captured the “baby” cue

With 2,859 characters of SVG markup, this is a balanced output – enough detail to be recognizable without being bloated.

View raw SVG source

```xml ```

11. nemotron-3-super:cloud

SVG size: 938 characters
Complexity: Compact
Shape elements: 7
Distinct colors: 5
Raw response: 1154 characters
Generation time: 12.1s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Baby/cute styling is present – the model captured the “baby” cue

With 938 characters of SVG markup, this is a compact output – minimal but potentially efficient for web embedding.

View raw SVG source

```xml ```

12. nemotron-3-ultra:cloud

SVG size: 5199 characters
Complexity: Medium
Shape elements: 25
Distinct colors: 8
Raw response: 6135 characters
Generation time: 236.2s

Analysis

Notable SVG techniques used:

• Uses <defs> blocks for reusable components (a sign of well-structured, editable SVG)
• Includes gradient fills for richer visual depth (great for ocean water and lighting)
• Applies SVG filters (e.g., drop shadows, blurs) for atmospheric effects
• Uses transforms (translate, rotate, scale) for positioning the shark and scene elements
• Contains XML comments, showing the model’s internal documentation of its work

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character
• Gills are present – the model captured anatomical detail
• Bubbles are present – the model created an underwater atmosphere
• Baby/cute styling is present – the model captured the “baby” cue

With 5,199 characters of SVG markup, this is a detailed output that balances richness with readability.

View raw SVG source

```xml ```

13. qwen3.5:397b-cloud

SVG size: 2023 characters
Complexity: Medium
Shape elements: 19
Distinct colors: 7
Raw response: 2203 characters
Generation time: 32.0s

Analysis

Notable SVG techniques used:

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

Concept adherence:

• Fin/dorsal fin is present – the model captured the shark’s defining feature
• Teeth/mouth is present – the model rendered the shark’s mouth
• Tail is present – the model captured the shark’s propulsion
• Eyes are present – the model gave the shark character

With 2,023 characters of SVG markup, this is a balanced output – enough detail to be recognizable without being bloated.

View raw SVG source

```xml ```

Conclusion

That’s all 13 baby shark SVGs from the active Ollama cloud models. The marine-life prompt tested a different skill set than our previous prompts: organic curves, underwater atmosphere, anatomical features, and cute character styling.

No single model is best at everything. A model that produces a stunning cherry blossom may produce a basic shark, and vice versa. The best model for you depends on your specific use case:

• For maximum detail: look at the “Very high” complexity models
• For speed: look at the generation times in each section
• For code quality: look at the raw SVG source – clean, well-structured code is easier to customize
• For concept adherence: check which models captured the shark’s fins, teeth, tail, and baby styling

Try the same prompt yourself with Ollama Cloud and see if you agree with our analysis. And check out the other posts in our SVG benchmark series:

• Duck on a bicycle
• Duck with a parachute
• Duck driving a jeep
• Cherry blossom trees
• Duck programmer debugging at 3am

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