LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Ministral 3 3B
◆ OPEN
by Mistral
OPEN WEIGHTSLOCALAPIPAIDSMALLEUROPE-MADE
RANKED #203 OF 316 ASSISTANTS · OVERALL #310 OF 493 RANKED · VIBE SCORE 5.8 · 76 VOICES ·
COPY BADGE pastes a README snippet with the rank SVG.
mistralai_Ministral-3-3B-Instruct-2512-GGUF on Hugging Face (image text to text). 19,672 downloads. Open weights for local or hosted use.
HAPPY10.0
14 mentions
SAD5.0
4 mentions
VIBE TRAJECTORY
26 weeks · 73 voices in this window
HOLDING UP
Recent vibe is steady. No clear fade in the crowd.
HOW TO USE IT
Open weights. You can use a hosted service, or download it and run it yourself, free.
RUN IT ON YOUR OWN MACHINE
Open weights: download Ministral 3 3B once and it's yours. No subscription, no rate limits, works offline.
3.8B total parameters (3.4B text + 0.4B vision), a Q4_K_M GGUF is about 2.4 GB
- BARE MINIMUM
Nvidia GeForce GTX 1650 4GB
4 GB VRAM
~$75 USEDQ4_K_M quant, ~30 tok/s, 8k context
- SWEET SPOT
Nvidia GeForce RTX 3060 12GB
12 GB VRAM
~$250 USEDQ8_0 or FP16, ~120 tok/s, 32k context with full vision offloading
- FULL POWER
Nvidia GeForce RTX 4070 12GB
12 GB VRAM
~$520 NEWBF16 / FP16 unquantized, ~225 tok/s, extended context window
LEARN HOW TO SET IT UP
WILL MY BOX RUN MINISTRAL 3 3B?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long Ministral 3 3B takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMNvidia GeForce GTX 1650 4GB | SWEET SPOTNvidia GeForce RTX 3060 12GB | FULL POWERNvidia GeForce RTX 4070 12GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~17 S★★★★★ | ~4 S★★★★★ | ~2 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~50 S★★★★★ | ~13 S★★★★★ | ~7 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~4.4 MIN★★★★★ | ~67 S★★★★★ | ~36 S★★★★★ |
| Build a backendan API with routes, storage and tests | ~14 MIN★★★★★ | ~3.5 MIN★★★★★ | ~1.9 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~8.3 MIN★★★★★ | ~2.1 MIN★★★★★ | ~67 S★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
People are mostly frustrated with Ministral 3 3B right now.
65 POSTS · 11 COMMENTS · REDDIT · GITHUB · LEMMY
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- Refuses or lectures more than people like (4 mentions)
TOPIC BY TOPIC
- Saying no too muchA SORE SPOT
0 thumbs up · 4 thumbs down
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with multimodal models like this one.
Grok 4.6
Qwen3.8 27B> SOURCES (12)
- REDDIT · 2025-12-02 · 583 UPVOTESIntroducing Mistral 3
Introducing Mistral 3
- REDDIT · 2026-10-18 · 160 UPVOTESlegal-it-8b: Ministral-3-8B fine-tuned for Italian law, trained on a European supercomputer
legal-it-8b: Ministral-3-8B fine-tuned for Italian law, trained on a European supercomputer
- REDDIT · 2026-02-23 · 96 UPVOTESI benchmarked 17 local LLMs on real MCP tool calling — single-shot AND agentic loop. The difference is massive.
I benchmarked 17 local LLMs on real MCP tool calling — single-shot AND agentic loop. The difference is massive.
- REDDIT · 2025-12-02 · 76 UPVOTESMistral 3 Family Released (10 Models): Large 3 hits 1418 Elo, Ministral 3 (3B/8B/14B) beats Qwen-VL. Full Benchmarks & Specs.
Mistral 3 Family Released (10 Models): Large 3 hits 1418 Elo, Ministral 3 (3B/8B/14B) beats Qwen-VL. Full Benchmarks & Specs.
- REDDIT · 2025-12-04 · 74 UPVOTESMistral Ministral 3 Reinforcement Learning is now in Unsloth! New RL sodoku example.
Mistral Ministral 3 Reinforcement Learning is now in Unsloth! New RL sodoku example.
- REDDIT · 2026-05-03 · 42 UPVOTESSide-by-side comparison of Qwen-Image, ERNIE Base/Turbo, and FLUX.2 Dev across 8 custom styles (single RTX 5090)
Side-by-side comparison of Qwen-Image, ERNIE Base/Turbo, and FLUX.2 Dev across 8 custom styles (single RTX 5090)
- REDDIT · 2026-10-05 · 36 UPVOTESMistral got one prompt and did it right.
Mistral got one prompt and did it right.
- REDDIT · 2026-02-23 · 34 UPVOTESMistral API quota and rate limits pools analysis for Free Tier plan (20.02.2026)
Mistral API quota and rate limits pools analysis for Free Tier plan (20.02.2026)
- REDDIT · 2025-12-04 · 26 UPVOTESHaving a great time with Ministral 3 3B 2512 Instruct
Having a great time with Ministral 3 3B 2512 Instruct
- REDDIT · 2026-10-07 · 21 UPVOTESA stock GLM-Edge-1.5B-Chat on a 4GB Galaxy A04e completed a real Amazon cart task
A stock GLM-Edge-1.5B-Chat on a 4GB Galaxy A04e completed a real Amazon cart task
- REDDIT · 2026-02-09 · 21 UPVOTESministral-3-3b is great model, give it a shot!
ministral-3-3b is great model, give it a shot!
- REDDIT · 2026-06-10 · 20 UPVOTESOllama Free Tier - Model status
Ollama Free Tier - Model status