Phi-3.5 Instruct: pricing, context window and benchmarks
Phi-3.5 Instruct is a open-weight model from Microsoft. It offers a 16,384-token context window, costs $2.67 per million input tokens and $14.62 per million output tokens, and averages 70.1 across the index's benchmark suite.
Pricing
Phi-3.5 Instruct costs $2.67 per million input tokens and $14.62 per million output tokens. A blended 3:1 input-to-output mix works out to about $5.66 per million tokens. Cached input is listed at $0.28 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $2.67 | per 1M tokens |
| Output | $14.62 | per 1M tokens |
| Cached input | $0.28 | per 1M tokens |
| Blended 3:1 | $5.66 | estimate per 1M tokens |
| 1M in + 1M out | $17.29 | 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 | 74.9 | 338 | 44th | |
| GPQA Diamond | 65.5 | 352 | 41th | |
| SWE-bench Verified | 56.1 | 376 | 37th | |
| HumanEval | 72.1 | 378 | 37th | |
| MATH-500 | 74.6 | 338 | 44th | |
| AIME 2025 | 66.6 | 323 | 46th | |
| LiveCodeBench | 59.6 | 373 | 38th | |
| IFEval | 79.2 | 353 | 41th | |
| TruthfulQA | 75.3 | 326 | 46th | |
| HellaSwag | 82.2 | 343 | 43th | |
| MMMU | 65.3 | 379 | 37th | |
| Arena Elo | 1346.8 | 369 | 39th |
What Phi-3.5 Instruct is good for
Phi-3.5 Instruct 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 $2.67 per million input tokens (hosted list price) and a 16,384-token context window, it is a practical option for teams that want more control than a closed API allows.
Strengths
- Input price of $2.67 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
- Context window of 16,384 tokens is small for long-document work.
- Average benchmark score is below the index median; test it on your own task.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Sonar Pro Mini | Perplexity | 8192 | $8.02 | 70.3 |
| Yi-Large Base | 01.AI | 16384 | $2.78 | 70.3 |
| Auto | OpenRouter | 32768 | $7.53 | 70.2 |
| Phi-3.5 Instruct | Microsoft | 16384 | $2.67 | 70.1 |
| Reka Core Mini | Reka AI | 8192 | $7.48 | 70.1 |
| Firesearch Max | Fireworks AI | 8192 | $7.80 | 69.9 |
FAQ
How much does Phi-3.5 Instruct cost?
Phi-3.5 Instruct is listed at $2.67 per million input tokens and $14.62 per million output tokens. Cached input is $0.28 per million tokens. A blended 3:1 input-to-output mix is about $5.66 per million tokens.
What is the context window of Phi-3.5 Instruct?
Phi-3.5 Instruct has a context window of 16,384 tokens and can emit up to 16,384 output tokens per request in this index.
Is Phi-3.5 Instruct open weights?
Yes. Phi-3.5 Instruct is listed as an open-weight model under a open weights licence.
How does Phi-3.5 Instruct score on benchmarks?
Phi-3.5 Instruct averages 70.1 across the scored benchmarks in this index, ranking #363 of 600 models. Its strongest relative result is Arena Elo at 1346.8.
What is Phi-3.5 Instruct best for?
Phi-3.5 Instruct 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 $2.67 per million input tokens (hosted list price) and a 16,384-token context window, it is a practical option for teams that want more control than a closed API allows.