Sources
Sites from which values are cited, and the publication licenses and terms checked.
- Source sites
- 13total
- Pages cited
- 54total
- Values cited
- 1150total
- With retrieval date and source excerpt
Source sites
| Name | URL | License and terms | Notes |
|---|---|---|---|
| Anthropic (announcement page and Claude documentation) | https://www.anthropic.com/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | Tables and the Accuracy vs Cost figure (each point's aria-label) on the announcement page, pricing table, model list, effort, and Claude Code model settings. Values for other vendor models are those appearing in the Anthropic announcement. |
| OpenAI (API documentation, Codex documentation, and Deployment Safety Hub) | https://developers.openai.com/ | Official vendor documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | Model list, pricing table, Codex model list, and system card (HealthBench table). The announcement page on openai.com could not be retrieved. |
| Google (Gemini model page, model card, and Gemini API documentation) | https://deepmind.google/models/gemini/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | Only the evaluation methodology PDF from deepmind-media is cited from storage.googleapis.com. The table includes rows where Google reproduced third-party measurement values (noted in conditions). |
| xAI (announcement page and API documentation) | https://x.ai/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | — |
| Z.ai (GLM documentation) | https://docs.z.ai/ | Official vendor documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | — |
| Moonshot AI (Kimi blog and platform) | https://kimi.com/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | — |
| Mistral AI (announcement page) | https://mistral.ai/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish scores and prices as facts with the source URL and a quote. | — |
| DeepSeek (API documentation) | https://api-docs.deepseek.com/ | Official vendor documentation. We have not confirmed a reuse license statement. We publish prices as facts with the source URL and a quote. | The benchmark table could not be retrieved because it was available only as an image, so there are no results. |
| Alibaba (Alibaba Cloud documentation and Alibaba Group announcement) | https://www.alibabacloud.com/ | Official vendor announcements and documentation. We have not confirmed a reuse license statement. We publish prices as facts with the source URL and a quote. | The official benchmark figures could not be retrieved, so there are no results. |
| Epoch AI(Capabilities & benchmarking) | https://epoch.ai/benchmarks | CC BY 4.0 (Epoch AI data; data collected externally is subject to its original license) | Only benchmarks run by Epoch AI itself from benchmark_data.zip are included. Attribution: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. This site selects and extracts Epoch AI’s in-house run data, normalizes model and metric names and score units and precision, and maps supported Model version suffixes to effort. We calculate ranks and percentiles within this site’s comparison groups and use the listed values in overall and role-based aggregations. |
| LMArena leaderboard dataset(Hugging Face) | https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset | CC BY 4.0 (license statement for the Hugging Face dataset) | The Hugging Face dataset lmarena-ai/leaderboard-dataset is used as the source, rather than the arena.ai site. This site selects and extracts values for the relevant models and categories from LMArena’s public dataset, normalizes model names and score precision, and maps supported model name variants to effort. We calculate ranks and percentiles within this site’s comparison groups and use the listed values in overall and role-based aggregations. |
| Official vendor model card documents on Hugging Face | https://huggingface.co/ | Official vendor model card documents (self-reported). We use only repositories whose LICENSE we have read in full and confirmed to have no non-commercial, geographic, or documentation restrictions, and publish scores as facts with the source URL and a quote. | Only values identifiable from the cards as the vendors' own evaluations are included. The model cards containing scores are moonshotai/Kimi-K3, moonshotai/Kimi-K3.1-Pro, and zai-org/GLM-5.3 (MIT-style license); deepseek-ai/DeepSeek-V4.1-Flash and deepseek-ai/DeepSeek-V4.1-Flash-1017 (MIT); Qwen/Qwen3.8-2.4T-A95B (Qwen3.8-Max License); and Qwen/Qwen3.8-27B (Apache License 2.0). The zai-org/GLM-5.3-Flash (MIT) card is referenced as a source of model information. |
| τ²-bench leaderboard (sierra-research/tau2-bench) | https://github.com/sierra-research/tau2-bench | MIT (Copyright (c) 2025 Sierra Research). Values and conditions are taken from the repository submission files (submission.json), and the description quotes text from schema.json and CHANGELOG.md in the same repository. The full copyright notice and permission notice are included in the "Data license notices" on this page. | Submission data for text models from the τ²-bench public leaderboard. The source type is determined by the source publisher, so all are third-party. Only values run and verified by Sierra (measured/reported by Sierra) are included. Model provider-declared or reported value entries (measured value entries from providers that appear on the leaderboard) are not included until their treatment in display and aggregation is decided. |
Pages cited
Data license notices
These data licenses require copyright and permission notices. They are reproduced in the original English.
τ²-bench leaderboard (sierra-research/tau2-bench)
MIT License
Copyright (c) 2025 Sierra Research
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
Fonts
Fonts bundled with this site (from npm @fontsource-variable packages 5.3.0). Japanese characters use fonts available in the viewing environment.
| Fonts | Copyright holder | License |
|---|---|---|
| Inter | Copyright 2016 The Inter Project Authors | SIL Open Font License 1.1 |
| Fraunces | Copyright 2020 The Fraunces Project Authors | SIL Open Font License 1.1 |