Zephyr Base: pricing, context window and benchmarks
Zephyr Base is a open-weight model from Hugging Face. It offers a 4,096-token context window, costs $1.26 per million input tokens and $6.45 per million output tokens, and averages 59.5 across the index's benchmark suite.
Pricing
Zephyr Base costs $1.26 per million input tokens and $6.45 per million output tokens. A blended 3:1 input-to-output mix works out to about $2.56 per million tokens. Cached input is listed at $0.18 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $1.26 | per 1M tokens |
| Output | $6.45 | per 1M tokens |
| Cached input | $0.18 | per 1M tokens |
| Blended 3:1 | $2.56 | estimate per 1M tokens |
| 1M in + 1M out | $7.71 | 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 | 65.4 | 447 | 26th | |
| GPQA Diamond | 57.2 | 438 | 27th | |
| SWE-bench Verified | 42.4 | 498 | 17th | |
| HumanEval | 64.8 | 443 | 26th | |
| MATH-500 | 62.6 | 440 | 27th | |
| AIME 2025 | 51.0 | 452 | 25th | |
| LiveCodeBench | 47.9 | 468 | 22th | |
| IFEval | 70.3 | 456 | 24th | |
| TruthfulQA | 65.1 | 458 | 24th | |
| HellaSwag | 73.1 | 480 | 20th | |
| MMMU | 54.2 | 506 | 16th | |
| Arena Elo | 1270.3 | 458 | 24th |
What Zephyr Base is good for
Zephyr 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 $1.26 per million input tokens (hosted list price) and a 4,096-token context window, it is a practical option for teams that want more control than a closed API allows.
Strengths
- Input price of $1.26 per million tokens is at or below the index median.
- Open weights make self-hosting, fine-tuning and data-residency review possible.
Trade-offs
- Context window of 4,096 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 |
|---|---|---|---|---|
| Qwen3-Coder 7B | Alibaba Cloud | 32768 | $1.24 | 59.6 |
| Llama 3.3 Groq 8B | Groq | 16384 | $1.17 | 59.5 |
| MiniMax-Text 8B | MiniMax | 4096 | $1.22 | 59.5 |
| Nova Lite Nano | Amazon | 8192 | $3.29 | 59.5 |
| QwQ 8B | Alibaba Cloud | 16384 | $1.37 | 59.5 |
| Zephyr Base | Hugging Face | 4096 | $1.26 | 59.5 |
FAQ
How much does Zephyr Base cost?
Zephyr Base is listed at $1.26 per million input tokens and $6.45 per million output tokens. Cached input is $0.18 per million tokens. A blended 3:1 input-to-output mix is about $2.56 per million tokens.
What is the context window of Zephyr Base?
Zephyr Base has a context window of 4,096 tokens and can emit up to 4,096 output tokens per request in this index.
Is Zephyr Base open weights?
Yes. Zephyr Base is listed as an open-weight model under a open weights licence.
How does Zephyr Base score on benchmarks?
Zephyr Base averages 59.5 across the scored benchmarks in this index, ranking #462 of 600 models. Its strongest relative result is Arena Elo at 1270.3.
What is Zephyr Base best for?
Zephyr 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 $1.26 per million input tokens (hosted list price) and a 4,096-token context window, it is a practical option for teams that want more control than a closed API allows.