Phi-3.5 Chat: pricing, context window and benchmarks
Phi-3.5 Chat is a open-weight model from Microsoft. It offers a 16,384-token context window, costs $2.64 per million input tokens and $12.13 per million output tokens, and averages 69.0 across the index's benchmark suite.
Pricing
Phi-3.5 Chat costs $2.64 per million input tokens and $12.13 per million output tokens. A blended 3:1 input-to-output mix works out to about $5.01 per million tokens. Cached input is listed at $0.31 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $2.64 | per 1M tokens |
| Output | $12.13 | per 1M tokens |
| Cached input | $0.31 | per 1M tokens |
| Blended 3:1 | $5.01 | estimate per 1M tokens |
| 1M in + 1M out | $14.77 | 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 | 74.4 | 345 | 43th | |
| GPQA Diamond | 64.9 | 362 | 40th | |
| SWE-bench Verified | 56.6 | 372 | 38th | |
| HumanEval | 71.0 | 389 | 35th | |
| MATH-500 | 69.0 | 388 | 35th | |
| AIME 2025 | 64.6 | 346 | 42th | |
| LiveCodeBench | 62.6 | 340 | 43th | |
| IFEval | 79.2 | 352 | 41th | |
| TruthfulQA | 72.4 | 377 | 37th | |
| HellaSwag | 80.9 | 373 | 38th | |
| MMMU | 63.7 | 393 | 35th | |
| Arena Elo | 1338.0 | 380 | 37th |
What Phi-3.5 Chat is good for
Phi-3.5 Chat 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 $2.64 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 $2.64 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 |
|---|---|---|---|---|
| Gemini 2.5 Pro Nano | Google DeepMind | 32768 | $7.28 | 69.2 |
| DBRX Base | Databricks | 32768 | $2.63 | 69.1 |
| Firesearch Reasoning | Fireworks AI | 32768 | $7.74 | 69.0 |
| Phi-3.5 Chat | Microsoft | 16384 | $2.64 | 69.0 |
| Claude 3.7 Sonnet Nano | Anthropic | 65536 | $6.93 | 68.9 |
| Arctic | Snowflake | 8192 | $2.72 | 68.8 |
FAQ
How much does Phi-3.5 Chat cost?
Phi-3.5 Chat is listed at $2.64 per million input tokens and $12.13 per million output tokens. Cached input is $0.31 per million tokens. A blended 3:1 input-to-output mix is about $5.01 per million tokens.
What is the context window of Phi-3.5 Chat?
Phi-3.5 Chat 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 Chat open weights?
Yes. Phi-3.5 Chat is listed as an open-weight model under a open weights licence.
How does Phi-3.5 Chat score on benchmarks?
Phi-3.5 Chat averages 69.0 across the scored benchmarks in this index, ranking #373 of 600 models. Its strongest relative result is Arena Elo at 1338.0.
What is Phi-3.5 Chat best for?
Phi-3.5 Chat 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 $2.64 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.