Ctrl K
Back to models

Mistral

Ministral 3 3B

  • Released
  • Open weights
  • Apache 2.0
Compare
Intelligence Index

Standard setting

4.8#82 of 83measured models
Measured · Artificial Analysis v4.3.2
API price from
$0.10 in / $0.10 out / 1M tokOpenRouter · Mistral
Context window
256K tok

Performance profile

ReasoningRank #100 of 107
35.8%
GPQA Diamond
CodingRank #92 of 94
15.3%
SciCode
Cost efficiencyRank #4 of 67
616.5pts / $
Intelligence ÷ eval cost

Capabilities

Context window

Modalities

Measured setting

Integration

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
Ministral 3 3B
Cost per task (USD) · log scale
Ministral 3 3BStandard
Intelligence Index
4.8
Cost / task
$0.0078
Evaluation time
57.7s

Pricing & access

API route
OpenRouter · MistralAPI
Tier & contextStandard

USD per 1M tokens

Input
$0.10
Output
$0.10
Cached input
$0.01

Run locally

GGUF
Context
Estimated memory

~4.8 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

16 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
2 GiB
Context cache
0.8 GiB16-bit cache
App allowance
2 GiB
Disk space
~9 GBIncludes image file
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · Q4_K_M · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/unsloth/Ministral-3-3B-Instruct-2512-GGUF:Q4_K_M
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter unsloth/Ministral-3-3B-Instruct-2512-GGUF, choose Q4_K_M, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

Images need extra memoryvLLM setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

Ministral 3 3B · Standard#106 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 103LFM2.5-2.6BReasoning6.2%
  6. 104Llama 3.2 Instruct 11B (Vision)Standard5.5%
  7. 105NVIDIA Nemotron Nano 12B v2 VLReasoning5.5%
  8. 106Ministral 3 3BStandard5.4%
  9. 106K2 Horizon 0.9BReasoning5.4%
  10. 108Tiny Aya GlobalStandard5.2%
  11. 109Granite 4.0 MicroStandard5.0%

Peers use their best measured result on this benchmark.

Alternatives

Each model at its highest-scoring measured setting · Intelligence Index v4.3.2 and its evaluation cost per task.