Small Models

Comparison of small open source LLMs (8b parameters or less)

General | Tool Usage | Chat Room | Logic | Vision | Vision OCR | Code Generation | More

All testing done with Ollama

	Key
✅	Pass
☑️	Partial Pass
❌	Fail
🕒	Timeout
😵‍💫	Hallucinations
🚫	Rejected prompt / unsafe content
💔	Malformed Response / Tool call

General

Tested with attogram/ollama-multirun

Small Model	Hovercraft	Pirate	Who
bakllava:7b	❌		✅
codellama:7b	✅	✅	✅
cogito:3b		✅	✅
cogito:bb		✅	✅
deepcoder:1.5b	✅	☑️	✅
deepseek-r1:1.5b	✅	☑️	✅
deepseek-r1:8b	☑️	✅	✅
dolphin-mistral:7b	☑️	✅	✅
dolphin3:8b	✅	✅	✅
gemma3:1b	☑️	✅	✅
gemma3:4b	☑️		✅
gemma:2b	☑️		✅
granite3.2-vision:2b	✅		✅
granite3.3:2b	☑️		✅
huihui_ai/baronllm-abliterated:8b	☑️		✅
llama3-groq-tool-use:8b	☑️		✅
llama3.2:1b	✅		☑️
llava-llama3:8b	☑️		✅
llava-phi3:3.8b	✅		✅
llava:7b	✅		✅
minicpm-v:8b	❌		✅
mistral:7b	☑️		✅
qwen2.5-coder:7b	✅		✅
qwen2.5vl:3b	✅		✅
qwen2.5vl:7b	☑️		✅
qwen3:1.7b	✅	🕒	✅
qwen3:4b	✅	🕒	✅
qwen3:8b	☑️	✅	✅
smollm2:135m	✅		✅
smollm2:360m	☑️		✅
smollm2:1.7b	✅		✅
stable-code:3b	☑️	☑️	✅
starcoder:7b	❌	❌	❌
Small Model	Hovercraft	Pirate	Who

deepseek-r1:7b
llama3.1:8b
llama3.3:3b
llama3:8b
phi:2.7b
phi3:3.8b
tinyllama:1.1b
qwen3:0.6b
qwen3:4b

^top

Tool Usage

Tested with attogram/ollama-bash-toolshed

Passing	Math	Ollama	Time	Web
cogito:3b	✅	✅	✅	✅
cogito:8b	✅	✅	✅	✅
granite3.3:2b	✅	✅	✅	✅
hermes3:8b	✅	✅	✅	✅
llama3.2:1b	✅	✅	✅	✅
llama3.2:3b	✅	✅	✅	✅
qwen3:1.7b	✅	✅	✅	✅
qwen3:4b	✅	✅	✅	✅
qwen3:8b	✅	✅	✅	✅

Failing	Math	Ollama	Time	Web
command-r7b:7b	❌	❌	❌	❌
granite3.2-vision:2b	❌	❌	❌	❌
granite3.3:8b	💔	💔	💔	💔
llama3-groq-tool-use:8b	✅	✅	❌	✅
mistral:7b	💔	💔	💔	💔
nemotron-mini:4b	✅	❌	✅	☑️
qwen2.5-coder:7b	💔	💔	💔	💔
qwen3:0.6b	✅	☑️	❌	✅
smollm2:1.7b	💔	💔	✅	😵‍💫

aya-expanse:8b
command-r7b-arabic:7b
granite3-dense:2b
granite3-dense:8b
granite3-moe:1b
granite3-moe:3b
granite3.1-dense:2b
granite3.1-dense:8b
granite3.1-moe:1b
granite3.1-moe:3b
granite3.2:2b
granite3.2:8b
hermes3:3b
llama3.1:8b
mixtral:8x7b
phi4-mini:3.8b
qwen2.5-coder:0.5b
qwen2.5-coder:1.5b
qwen2.5-coder:3b
qwen2.5:0.5b
qwen2.5:1.5b
qwen2.5:3b
qwen2.5:7b
qwen2:0.5b
qwen2:1.5b
qwen2:7b

Hituzip/gemma3-tools:4b
PetrosStav/gemma3-tools:4b
hhao/qwen2.5-coder-tools:0.5b
hhao/qwen2.5-coder-tools:1.5b
hhao/qwen2.5-coder-tools:3b
hhao/qwen2.5-coder-tools:7b
ishumilin/deepseek-r1-coder-tools-tuned:1.5b
ishumilin/deepseek-r1-coder-tools-tuned:7b
ishumilin/deepseek-r1-coder-tools-tuned:8b
ishumilin/deepseek-r1-coder-tools:1.5b
ishumilin/deepseek-r1-coder-tools:7b
ishumilin/deepseek-r1-coder-tools:8b
maryasov/qwen2.5-coder-cline:7b
orieg/gemma3-tools:1b
orieg/gemma3-tools:4b
tom_himanen/deepseek-r1-roo-cline-tools:1.5b
tom_himanen/deepseek-r1-roo-cline-tools:7b
tom_himanen/deepseek-r1-roo-cline-tools:8b

^top

Chat Room

Tested with attogram/llm-council

Small Model	Peace	Quit	Topic
bakllava:7b		❌
codellama:7b		✅
deepcoder:1.5b		❌
deepseek-r1:1.5b		✅	❌
deepseek-r1:8b		✅
dolphin-mistral:7b		✅
dolphin3:8b	✅	✅
gemma3:1b		✅
gemma3:4b	✅	✅	✅
gemma:2b		✅
granite3.2-vision:2b		❌
granite3.3:2b	✅	✅	❌
huihui_ai/baronllm-abliterated:8b		✅
llama3-groq-tool-use:8b		✅
llama3.2:1b		❌
llava-llama3:8b		✅
llava-phi3:3.8b		✅
llava:7b		✅
minicpm-v:8b	✅	✅
mistral:7b	✅	✅
qwen2.5-coder:7b		✅
qwen2.5vl:3b		✅
qwen2.5vl:7b		✅
qwen3:1.7b		✅
qwen3:8b	✅	✅
smollm2:135m		❌
smollm2:360m		❌
smollm2:1.7b		❌
stable-code:3b		❌
starcoder:7b		❌
Small Model	Peace	Quit	Topic

^top

Logic

Tested with attogram/ollama-multirun

Small Model	Socrates	Strawberry
bakllava:7b	✅	❌
codellama:7b	❌	❌
deepcoder:1.5b	✅	✅
deepseek-r1:1.5b	✅	✅
deepseek-r1:8b	✅	🕒
dolphin-mistral:7b	✅	❌
dolphin3:8b	✅	❌
gemma3:1b	✅	✅
gemma3:4b	✅	✅
gemma:2b	✅	✅
granite3.2-vision:2b	✅	✅
granite3.3:2b	✅	✅
huihui_ai/baronllm-abliterated:8b	✅	❌
llama3-groq-tool-use:8b	✅	✅
llama3.2:1b	❌	❌
llava-llama3:8b	✅	❌
llava-phi3:3.8b	✅	❌
llava:7b	✅	❌
minicpm-v:8b	✅	❌
mistral:7b	✅	❌
qwen2.5-coder:7b	✅	❌
qwen2.5vl:3b	✅	✅
qwen2.5vl:7b	✅	❌
qwen3:1.7b	✅	✅
qwen3:8b	✅	✅
smollm2:135m	☑️	❌
smollm2:360m	✅	❌
smollm2:1.7b	✅	❌
stable-code:3b	❌	❌
starcoder:7b	❌	❌
Small Model	Socrates	Strawberry

^top

Vision

Tested with attogram/ollama-multirun

Passing	Describe	Keywords	Faces
bakllava:7b	✅	❌	❌
granite3.2-vision:2b	✅	✅	❌
minicpm-v:8b	✅	✅	☑️
qwen2.5vl:3b	✅	✅	❌
qwen2.5vl:7b	✅	✅	✅
gemma3:4b	✅	✅	✅
llava:7b	✅	✅	✅
llava-llama3:8b	✅	☑️	☑️
llava-phi3:3.8b	✅	☑️	☑️
moondream:1.8b	✅	❌	❌

^top

Vision OCR

Tested with attogram/ollama-multirun

Passing	OCR
granite3.2-vision:2b	☑️
minicpm-v:8b	☑️
qwen2.5vl:3b	✅
qwen2.5vl:7b	✅

Failing	OCR
bakllava:7b	❌
gemma3:4b	🕒
llava:7b	❌
llava-llama3:8b	❌
llava-phi3:3.8b	❌
moondream:1.8b	❌

^top

Code Generation

Small Model	bash-1	js-1	php-1	python-1
bakllava:7b	❌		☑️
codellama:7b	✅ ️		✅
deepcoder:1.5b	❌		✅
deepseek-r1:1.5b	❌		✅
deepseek-r1:8b	🕒		🕒
dolphin-mistral:7b	✅		✅
dolphin3:8b	✅		✅
gemma3:1b	☑️		☑️
gemma3:4b	☑️		✅
gemma:2b	✅		❌
granite3.2-vision:2b	✅		✅
granite3.3:2b	☑️		✅
huihui_ai/baronllm-abliterated:8b	✅		✅
llama3-groq-tool-use:8b	✅		✅
llama3.2:1b	☑️		☑️
llava-llama3:8b	☑️		✅
llava-phi3:3.8b	✅		✅
llava:7b	✅		✅
minicpm-v:8b	❌		✅
mistral:7b	☑️		✅
qwen2.5-coder:7b	✅		✅
qwen2.5vl:3b	☑️		✅
qwen2.5vl:7b	✅		✅
qwen3:1.7b	✅		✅
qwen3:8b	✅		✅
smollm2:135m	❌		❌
smollm2:360m	☑️		☑️
smollm2:1.7b	✅		☑️
stable-code:3b	☑️		✅
starcoder:7b	☑️		❌
Small Model	bash-1	js-1	php-1	python-1

^top

More AI from the Attogram Project

Project	Github Repo	Description
Multirun	ollama-multirun	Run a prompt against all, or some, of your models running on Ollama. Creates web pages with the output, performance statistics and model info. All in a single Bash shell script.
Toolshed	ollama-bash-toolshed	Chat with tool calling models. Sample tools included. Add new tools to your shed with ease. Runs on Ollama. All via Bash shell scripts.
LLM Council	llm-council	Start a chat room between all, or some, of your models running on Ollama. All in a single Bash shell script.
Ollama Bash Lib	ollama-bash-lib	A Bash Library to interact with Ollama
Small Models	small-models	Comparison of small open source LLMs (8b parameters or less)
AI Test Zone	ai_test_zone	Test results hosted on https://attogram.github.io/ai_test_zone/

^top