Pythia Instruct: pricing, context window and benchmarks
Pythia Instruct is a open-weight model from EleutherAI. It offers a 8,192-token context window, costs $1.41 per million input tokens and $5.70 per million output tokens, and averages 60.8 across the index's benchmark suite.
Pricing
Pythia Instruct costs $1.41 per million input tokens and $5.70 per million output tokens. A blended 3:1 input-to-output mix works out to about $2.48 per million tokens. Cached input is listed at $0.14 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $1.41 | per 1M tokens |
| Output | $5.70 | per 1M tokens |
| Cached input | $0.14 | per 1M tokens |
| Blended 3:1 | $2.48 | estimate per 1M tokens |
| 1M in + 1M out | $7.11 | 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 | 65.3 | 448 | 25th | |
| GPQA Diamond | 58.3 | 424 | 29th | |
| SWE-bench Verified | 47.1 | 445 | 26th | |
| HumanEval | 66.3 | 427 | 29th | |
| MATH-500 | 63.5 | 429 | 29th | |
| AIME 2025 | 52.3 | 443 | 26th | |
| LiveCodeBench | 46.8 | 480 | 20th | |
| IFEval | 72.4 | 428 | 29th | |
| TruthfulQA | 66.7 | 435 | 28th | |
| HellaSwag | 73.8 | 459 | 24th | |
| MMMU | 56.2 | 485 | 19th | |
| Arena Elo | 1291.1 | 430 | 28th |
What Pythia Instruct is good for
Pythia 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 $1.41 per million input tokens (hosted list price) and a 8,192-token context window, it is a practical option for teams that want more control than a closed API allows.
Strengths
- Input price of $1.41 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 8,192 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 |
|---|---|---|---|---|
| Llama 4 Scout 8B | Meta AI | 8192 | $1.46 | 61.2 |
| Mistral Large 3 7B | Mistral AI | 32768 | $1.44 | 61.1 |
| Reka Flash Nano | Reka AI | 16384 | $3.93 | 61.0 |
| MiniMax-M2 7B | MiniMax | 8192 | $1.44 | 60.8 |
| Pythia Instruct | EleutherAI | 8192 | $1.41 | 60.8 |
| SmolLM3 Base | Hugging Face | 16384 | $1.38 | 60.4 |
FAQ
How much does Pythia Instruct cost?
Pythia Instruct is listed at $1.41 per million input tokens and $5.70 per million output tokens. Cached input is $0.14 per million tokens. A blended 3:1 input-to-output mix is about $2.48 per million tokens.
What is the context window of Pythia Instruct?
Pythia Instruct has a context window of 8,192 tokens and can emit up to 8,192 output tokens per request in this index.
Is Pythia Instruct open weights?
Yes. Pythia Instruct is listed as an open-weight model under a open weights licence.
How does Pythia Instruct score on benchmarks?
Pythia Instruct averages 60.8 across the scored benchmarks in this index, ranking #446 of 600 models. Its strongest relative result is Arena Elo at 1291.1.
What is Pythia Instruct best for?
Pythia 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 $1.41 per million input tokens (hosted list price) and a 8,192-token context window, it is a practical option for teams that want more control than a closed API allows.