Pythia: pricing, context window and benchmarks
Pythia is a open-weight model from EleutherAI. It offers a 32,768-token context window, costs $1.29 per million input tokens and $5.47 per million output tokens, and averages 59.8 across the index's benchmark suite.
Pricing
Pythia costs $1.29 per million input tokens and $5.47 per million output tokens. A blended 3:1 input-to-output mix works out to about $2.33 per million tokens. Cached input is listed at $0.19 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $1.29 | per 1M tokens |
| Output | $5.47 | per 1M tokens |
| Cached input | $0.19 | per 1M tokens |
| Blended 3:1 | $2.33 | estimate per 1M tokens |
| 1M in + 1M out | $6.76 | 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 | 63.1 | 487 | 19th | |
| GPQA Diamond | 55.6 | 455 | 24th | |
| SWE-bench Verified | 44.8 | 467 | 22th | |
| HumanEval | 62.8 | 460 | 23th | |
| MATH-500 | 63.6 | 428 | 29th | |
| AIME 2025 | 50.0 | 460 | 23th | |
| LiveCodeBench | 52.2 | 427 | 29th | |
| IFEval | 70.4 | 454 | 24th | |
| TruthfulQA | 67.2 | 426 | 29th | |
| HellaSwag | 72.8 | 485 | 19th | |
| MMMU | 55.0 | 497 | 17th | |
| Arena Elo | 1259.9 | 476 | 21th |
What Pythia is good for
Pythia 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 $1.29 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 $1.29 per million tokens is at or below the index median.
- Open weights make self-hosting, fine-tuning and data-residency review possible.
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 |
|---|---|---|---|---|
| Llama 4 Scout 7B | Meta AI | 32768 | $1.30 | 60.0 |
| Kimi K1.5 7B | Moonshot AI | 8192 | $1.35 | 59.9 |
| Pythia | EleutherAI | 32768 | $1.29 | 59.8 |
| Qwen3 7B | Alibaba Cloud | 32768 | $1.37 | 59.8 |
| GPT-NeoX Base | EleutherAI | 16384 | $1.22 | 59.7 |
| Qwen3-Coder 7B | Alibaba Cloud | 32768 | $1.24 | 59.6 |
FAQ
How much does Pythia cost?
Pythia is listed at $1.29 per million input tokens and $5.47 per million output tokens. Cached input is $0.19 per million tokens. A blended 3:1 input-to-output mix is about $2.33 per million tokens.
What is the context window of Pythia?
Pythia has a context window of 32,768 tokens and can emit up to 16,384 output tokens per request in this index.
Is Pythia open weights?
Yes. Pythia is listed as an open-weight model under a open weights licence.
How does Pythia score on benchmarks?
Pythia averages 59.8 across the scored benchmarks in this index, ranking #454 of 600 models. Its strongest relative result is Arena Elo at 1259.9.
What is Pythia best for?
Pythia 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 $1.29 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.