Phi-4 Base: pricing, context window and benchmarks
Phi-4 Base is a open-weight model from Microsoft. It offers a 32,768-token context window, costs $3.15 per million input tokens and $11.39 per million output tokens, and averages 71.2 across the index's benchmark suite.
Pricing
Phi-4 Base costs $3.15 per million input tokens and $11.39 per million output tokens. A blended 3:1 input-to-output mix works out to about $5.21 per million tokens. Cached input is listed at $0.35 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $3.15 | per 1M tokens |
| Output | $11.39 | per 1M tokens |
| Cached input | $0.35 | per 1M tokens |
| Blended 3:1 | $5.21 | estimate per 1M tokens |
| 1M in + 1M out | $14.54 | estimate |
Benchmarks
Scores below are modelled estimates on a 0–100 scale. Use them to compare relative position, not as official vendor results.
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| MMLU-Pro | 74.0 | 354 | 41th | |
| GPQA Diamond | 64.9 | 363 | 40th | |
| SWE-bench Verified | 62.7 | 262 | 56th | |
| HumanEval | 73.9 | 359 | 40th | |
| MATH-500 | 71.9 | 368 | 39th | |
| AIME 2025 | 64.6 | 347 | 42th | |
| LiveCodeBench | 63.7 | 329 | 45th | |
| IFEval | 81.2 | 302 | 50th | |
| TruthfulQA | 74.4 | 346 | 42th | |
| HellaSwag | 84.8 | 258 | 57th | |
| MMMU | 67.0 | 359 | 40th | |
| Arena Elo | 1373.8 | 299 | 50th |
What Phi-4 Base is good for
Phi-4 Base is an open-weight model, so it can be self-hosted or run through a third-party inference provider. That matters for data residency, fine-tuning and long-run cost control. At $3.15 per million input tokens (hosted list price) and a 32,768-token context window, it is a practical option for teams that want more control than a closed API allows.
Strengths
- Input price of $3.15 per million tokens is at or below the index median.
- Open weights make self-hosting, fine-tuning and data-residency review possible.
- Accepts more than text, including image input where the provider supports it.
Trade-offs
- Average benchmark score is below the index median; test it on your own task.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Auto Max | OpenRouter | 8192 | $8.41 | 71.2 |
| Auto Pro | OpenRouter | 8192 | $8.41 | 71.2 |
| Phi-4 Base | Microsoft | 32768 | $3.15 | 71.2 |
| Fusion | OpenRouter | 16384 | $8.41 | 71.1 |
| Gemma 3 Base | Google DeepMind | 65536 | $3.09 | 71.1 |
| Mixtral Base | Mistral AI | 65536 | $2.95 | 71.1 |
FAQ
How much does Phi-4 Base cost?
Phi-4 Base is listed at $3.15 per million input tokens and $11.39 per million output tokens. Cached input is $0.35 per million tokens. A blended 3:1 input-to-output mix is about $5.21 per million tokens.
What is the context window of Phi-4 Base?
Phi-4 Base has a context window of 32,768 tokens and can emit up to 32,768 output tokens per request in this index.
Is Phi-4 Base open weights?
Yes. Phi-4 Base is listed as an open-weight model under a open weights licence.
How does Phi-4 Base score on benchmarks?
Phi-4 Base averages 71.2 across the scored benchmarks in this index, ranking #342 of 600 models. Its strongest relative result is Arena Elo at 1373.8.
What is Phi-4 Base best for?
Phi-4 Base is an open-weight model, so it can be self-hosted or run through a third-party inference provider. That matters for data residency, fine-tuning and long-run cost control. At $3.15 per million input tokens (hosted list price) and a 32,768-token context window, it is a practical option for teams that want more control than a closed API allows.