vs
0:1
overall capability wins (2 ties across 3 scored). Phi-4-mini leads overall.
- Pricing0:3
- Speed0:2
- Specs0:1
Verdict
Phi-4-mini allows ~4.0x more max output (16K vs 4K); Phi-4-mini is ~1.7x cheaper on a blended token basis than Phi-3.5-mini.
- Phi-4-mini allows ~4.0x more max output (16K vs 4K)
- Phi-4-mini is ~1.7x cheaper on a blended token basis than Phi-3.5-mini
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Phi-3.5-mini
- $0.26
- Phi-4-mini
- $0.15
Phi-4-mini is about 1.7x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Microsoft
Microsoft
Family
Phi
Phi
License
MIT
MIT
Open weights
Yes
Yes
Release
Aug 21, 2024
Feb 1, 2025
Knowledge cutoff
-
-
API / provider
Azure AI Foundry
Azure AI Foundry
Modalities
text → text
text → text
Specs
Context window
131K
131K
Max output
4K
16K
Parameters
3.8B
3.8B
Pricing
Input $/1M
$0.13
$0.07
Output $/1M
$0.52
$0.30
Blended $/1M (3∶1)
$0.23
$0.13
Speed
tok/s
130
140
TTFT (s)
0.2
0.18
Reasoning
MMLU-Pro
—
52.8
GPQA Diamond
—
33.1
Humanity's Last Exam
—
4.5
AIME 2025
—
6.7
MATH-500
—
69.6
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
9.5
AA-LCR v1.1
—
15.3
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
21.1
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
12.6
Terminal-Bench 2.1
—
0.4
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
10.8
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Phi-3.5-mini | Phi-4-mini | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Microsoft | Microsoft | — |
| Family | Phi | Phi | — |
| License | MIT | MIT | — |
| Open weights | Yes | Yes | — |
| Release | Aug 21, 2024 | Feb 1, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Azure AI Foundry | Azure AI Foundry | — |
| Modalities | text → text | text → text | — |
| Specs | |||
| Context window | 131K | 131K | tie |
| Max output | 4K | 16K | B +12K |
| Parameters | 3.8B | 3.8B | tie |
| Pricing | |||
| Input $/1M | $0.13 | $0.07 | B +$0.06 |
| Output $/1M | $0.52 | $0.30 | B +$0.22 |
| Blended $/1M (3∶1) | $0.23 | $0.13 | B +$0.10 |
| Speed | |||
| tok/s | 130 | 140 | B +10 |
| TTFT (s) | 0.2 | 0.18 | B +0.02 |
| Reasoning | |||
| MMLU-Pro | — | 52.8 | — |
| GPQA Diamond | — | 33.1 | — |
| Humanity's Last Exam | — | 4.5 | — |
| AIME 2025 | — | 6.7 | — |
| MATH-500 | — | 69.6 | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 9.5 | — |
| AA-LCR v1.1 | — | 15.3 | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | 21.1 | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | 12.6 | — |
| Terminal-Bench 2.1 | — | 0.4 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | 10.8 | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Phi-3.5-mini
- Context
- 131K / 4K out
- Parameters
- 3.8B
- Price
- $0.13 / $0.52
- Speed
- 130 tok/s · 0.2s TTFT
- Modalities
- text → text
- License
- MIT
Prior Phi SLM generation — still common on Azure for low-cost classification before the Phi-4 refresh.
Phi-4-mini
- Context
- 131K / 16K out
- Parameters
- 3.8B
- Price
- $0.07 / $0.30
- Speed
- 140 tok/s · 0.18s TTFT
- Modalities
- text → text
- License
- MIT
Next-gen Microsoft SLM with grouped-query attention and built-in function calling — strong edge and mobile deployment choice.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.
These two models have no overlapping published benchmarks in our dataset. Compare specs and pricing instead.