whatmodel.sh / Bielik 4.5B v3.0 Instruct

Bielik 4.5B v3.0 Instruct

+ contribute

Bielik 4.5B v3.0 Instruct is an open-weights language model from SpeakLeash, released April 2025. Its published size is 4.76B parameters. Licence: apache-2.0.

Catalogue identity notes

Maker repo speakleash/Bielik-4.5B-v3.0-Instruct @4b1220a: licence tag apache-2.0. 4,757,260,288 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2025-04: month the maker's Hugging Face repo was created (createdAt 2025-04-18). Gated repository (licence acceptance required on Hugging Face).

47.5±2.2#391 of 719 LLMsfrom 5 benchmark results across 4 sources

$ how it compares

The LLMs ranked either side of Bielik 4.5B v3.0 Instruct, out of 828 with a score, then more from SpeakLeash.

#	model	score	where	RAM	released
466	DeepSeek-Coder 6.7B Instruct DeepSeek	48±1	local	~5.8GB	2023-10
467	MiniCPM3-4B OpenBMB	48±2	local	~4GB	2024-09
468	Bielik 4.5B v3.0 Instruct SpeakLeash	48±2	local	~4.3GB	2025-04
469	Qwen3.5 2B Alibaba	47±1	local	~2.9GB	2026-03
470	Gemma 2 9B IT Google	47±1	local	~7GB	2024-06
	more from SpeakLeash
325	Bielik 11B v3.0 Instruct SpeakLeash	55±2	local	~8.1GB	2025-11
338	Bielik 11B v2.6 Instruct SpeakLeash	54±3	local	~8.1GB	2025-06
371	Bielik 11B v2.1 Instruct SpeakLeash	53±3	local	~8.5GB	2024-08

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests5 not in a task · 98% of its score

test	result	vs the others in its run	difficulty	counts	why
CPTUB average (SpeakLeash, 0-5 judge scale) lab	67.5% ✎	▁▂▂▂▃█▃▆▅▄▂▃ beats 49/95	easy · avg 66%	▇▇▇▇▇▇▇▇▇▇▇▇▇ 32%	at its level
Polish medical board exams, bias-mitigated MCQ protocol (7-exam average)	37.8% ✎	▄·▂▃▂··█▂▄·▄ beats 3/13	hard · avg 48%	▇▇▇▇▇▇▇▇▇▇▇ 27%	at its level
Polish EQ-Bench v2 (SpeakLeash) lab	56.2% ✎	▁··▁▁▂▂▆▅▆█▆ beats 32/70	medium · avg 64%	▇▇▇▇▇▇▇▇▇▇ 26%	at its level
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy	72.2% ✎	▃·▁▁▂▁▂▅█▇▄▁ beats 56/100	medium · avg 68%	▇▇▇ 8.0%	at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	80.5% ✎	▁▁▁▁▁▁▁▂▄█▄▃ beats 69/146	easy · avg 83%	▇▇ 5.6%	at its level · noisy

$ where to run it

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run Bielik 4.5B v3.0 Instruct · needs ~2.9 GB · from huggingface.co/second-state
chat in the terminal
llama cli -hf second-state/Bielik-4.5B-v3.0-Instruct-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf second-state/Bielik-4.5B-v3.0-Instruct-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Bielik 4.5B v3.0 Instruct · needs ~2.9 GB · from huggingface.co/second-state
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/second-state/Bielik-4.5B-v3.0-Instruct-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Bielik 4.5B v3.0 Instruct · needs ~2.9 GB · from huggingface.co/second-state
chat in the terminal
ollama run hf.co/second-state/Bielik-4.5B-v3.0-Instruct-GGUF:Q4_K_M

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Bielik 4.5B v3.0 Instruct · needs ~2.7 GB, Mac only · from huggingface.co/futurist-ai
chat in the terminal
mlx_lm.chat --model futurist-ai/Bielik-4.5B-v3.0-Instruct-MLX-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model futurist-ai/Bielik-4.5B-v3.0-Instruct-MLX-4bit

Other ways to run it

Run it as published · needs ~9.51 GB · from huggingface.co/speakleash (official)
uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='speakleash/Bielik-4.5B-v3.0-Instruct', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/speakleash ↗

$ how the score was worked out · SpeakLeash CPTUB (Complex Pol… 32%, Reassessing High-Performing L… 27%, SpeakLeash Polish EQ-Bench le… 26% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark) · lab-reported	1	▇▇▇▇▇▇▇▇▇▇ 32%	MIT
Reassessing High-Performing LLMs on Polish Medical Exams: True Competence or Bias-Driven Performance? (arXiv:2606.12250v1)	1	▇▇▇▇▇▇▇▇ 27%	CC BY-SA 4.0
SpeakLeash Polish EQ-Bench leaderboard · lab-reported	1	▇▇▇▇▇▇▇▇ 26%	Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard	2	▇▇▇▇ 14%	MIT
prior · pull toward the middle		▇ 2%
weighted average		= 47.5

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json