The AI arena is free today

Open Superagent

Ministral 3 (3B Instruct 2512) vs Phi 4

Ministral 3 (3B Instruct 2512) and Phi 4 are closely matched at 2.2 and 5.4 on the LLM Stats Score.

Mistral AI · Microsoft · Updated for 2026

Which is better?

Ministral 3 (3B Instruct 2512) and Phi 4 are closely matched on the overall LLM Stats Score at 2.2 and 5.4.

The models split the 2 individual benchmarks reported for both models evenly.

Based on current LLM Stats indexes, shared benchmarks, pricing, and model metadata for 2026.

Choose Ministral 3 (3B Instruct 2512)

  • you want the most recent training data — it shipped Dec 2025

Choose Phi 4

  • you want predictable pricing at $0.07/M input and $0.14/M output

At a glance

The differences that matter most.

Core performance indexes
2.2
#324
5.4
#304
2.7
#313
6.5
#295
Cost, coverage & limits
Benchmark wins
1 of 2
1 of 2
Input price
— / M
$0.07 / M
Output price
— / M
$0.14 / M
Context window
—
16,384

Individual benchmarks

4 reported for Ministral 3 (3B Instruct 2512) · 13 for Phi 4

2 shared

Ministral 3 (3B Instruct 2512) outperforms in 1 benchmarks (MATH), while Phi 4 is better at 1 benchmark (Arena Hard).

Both models are evenly matched across the benchmarks.

Mon Oct 05 2026 • llm-stats.com

Human preference

Blind head-to-head votes and playground preference scores

Model Size

Parameter count comparison

11.7B diff

Phi 4 has 11.7B more parameters than Ministral 3 (3B Instruct 2512), making it 390.0% larger.

Mistral AI
Ministral 3 (3B Instruct 2512)
3.0Bparameters
Microsoft
Phi 4
14.7Bparameters
3.0B
Ministral 3 (3B Instruct 2512)
14.7B
Phi 4

Context Window

Maximum input and output token capacity

Only Phi 4 specifies input context (16,384 tokens). Only Phi 4 specifies output context (16,384 tokens).

Mistral AI
Ministral 3 (3B Instruct 2512)
Input- tokens
Output- tokens
Microsoft
Phi 4
Input16,384 tokens
Output16,384 tokens
Mon Oct 05 2026 • llm-stats.com

Input capabilities

Documented input modalities across available providers

Ministral 3 (3B Instruct 2512) supports multimodal inputs, whereas Phi 4 does not.

Ministral 3 (3B Instruct 2512) can handle both text and other forms of data like images, making it suitable for multimodal applications.

Ministral 3 (3B Instruct 2512)

Text
Images
Audio
Video

Phi 4

Text
Images
Audio
Video

License

Usage and distribution terms

Ministral 3 (3B Instruct 2512) is licensed under Apache 2.0, while Phi 4 uses MIT.

License differences may affect how you can use these models in commercial or open-source projects.

Ministral 3 (3B Instruct 2512)

Apache 2.0

Open weights

Phi 4

MIT

Open weights

Release Timeline

When each model was launched

Ministral 3 (3B Instruct 2512) was released on 2025-12-04, while Phi 4 was released on 2024-12-12.

Ministral 3 (3B Instruct 2512) is 12 months newer than Phi 4.

Ministral 3 (3B Instruct 2512)

Dec 4, 2025

10 months ago

11mo newer
Phi 4

Dec 12, 2024

1.8 years ago

Knowledge Cutoff

When training data ends

Phi 4 has a documented knowledge cutoff of 2024-06-01, while Ministral 3 (3B Instruct 2512)'s cutoff date is not specified.

We can confirm Phi 4's training data extends to 2024-06-01, but cannot make a direct comparison without Ministral 3 (3B Instruct 2512)'s cutoff date.

Ministral 3 (3B Instruct 2512)

—

Phi 4

Jun 2024

Outputs Comparison

Notice missing or incorrect data?

Judge for yourself.

Run your own prompts against Ministral 3 (3B Instruct 2512) and Phi 4 side-by-side, then vote on the output you prefer.

Ministral 3 (3B Instruct 2512)
✓ Preferred
Phi 4
Open in Playground

FAQ

Common questions about Ministral 3 (3B Instruct 2512) vs Phi 4.

Which is better, Ministral 3 (3B Instruct 2512) or Phi 4?

Ministral 3 (3B Instruct 2512) and Phi 4 are closely matched on the LLM Stats Score at 2.2 and 5.4. Ministral 3 (3B Instruct 2512) is made by Mistral AI and Phi 4 is made by Microsoft. The best choice depends on your use case — compare their capability indexes, individual benchmarks, pricing, and limits above.

How does Ministral 3 (3B Instruct 2512) compare to Phi 4 in benchmarks?

Ministral 3 (3B Instruct 2512) scores MATH: 83.0%, Wild Bench: 56.8%, Arena Hard: 30.5%, MM-MT-Bench: 7.8%. Phi 4 scores MMLU: 84.8%, HumanEval+: 82.8%, HumanEval: 82.6%, MGSM: 80.6%, MATH: 80.4%.

What are the context window sizes for Ministral 3 (3B Instruct 2512) and Phi 4?

Ministral 3 (3B Instruct 2512) supports an unknown number of tokens and Phi 4 supports 16K tokens. A larger context window lets you process longer documents, conversations, or codebases in a single request.

What are the main differences between Ministral 3 (3B Instruct 2512) and Phi 4?

Key differences include LLM Stats Score (2.2 vs 5.4), multimodal support (yes vs no), licensing (Apache 2.0 vs MIT). See the full comparison above for benchmark-by-benchmark results.

Who makes Ministral 3 (3B Instruct 2512) and Phi 4?

Ministral 3 (3B Instruct 2512) is developed by Mistral AI and Phi 4 is developed by Microsoft.