从语言到影像，从开源到 API。读懂模型能力、比较真实成本，
为你的下一个想法，找到合适的智能引擎。
> **Join the Discord** for updates, roadmaps, projects, or just to chat. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities — fully functional, 100% of what the original authors intended, just without the refusals. Built from the official QAT weights, so the 4-bit quant stays close to full-precision quality. The **Balanced** variant (recommended — 99%+ of users will be happy here) uses optimized full uncensoring tuned especially for agentic coding, reasoning, creative writing and reliability-critical tasks. It reasons before answering and stays dependable and on-instruction. An **Aggressive** variant, for cases where Balanced still deflects too much, after current testing is not required.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities — fully functional, 100% of what the original authors intended, just without the refusals. Built from the official QAT weights, so the 4-bit quant stays close to full-precision quality. The **Balanced** variant (recommended — 99%+ of users will be happy here) uses optimized full uncensoring tuned especially for agentic coding, reasoning, creative writing and reliability-critical tasks. It reasons before answering and stays dependable and on-instruction. An **Aggressive** variant, for cases where Balanced still deflects too much, after current testing is not required.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. Gemma4-26B-A4B uncensored by HauhauCS. **0/465 Refusals*** **Release Candidate after over 1 month of nonstop work on this one.** > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities — fully functional, 100% of what the original authors intended, just without the refusals. Built from the official QAT weights, so the 4-bit quant stays close to full-precision quality. The **Balanced** variant (recommended — 99%+ of users will be happy here) uses optimized full uncensoring tuned especially for agentic coding, reasoning, creative writing and reliability-critical tasks. It reasons before answering and stays dependable and on-instruction. An **Aggressive** variant, for cases where Balanced still deflects too much, after current testing is not required.
译文排队中 · 当前显示来源原文
No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. The Aggressive variant removes more refusal behavior. Use this if the Balanced variant still refuses too much. **Important:** - Disable repeat penalty (or `--repeat-penalty 1.0`) - For llama.cpp: use `--min-p 0.01` (default 0.05 is too high) - Use `--jinja` flag for llama.cpp
译文排队中 · 当前显示来源原文
No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. If you're doing agentic coding, use the Balanced variants. Good balance between capability and not refusing everything. **Important:** - Disable repeat penalty (or `--repeat-penalty 1.0`) - For llama.cpp: use `--min-p 0.01` (default 0.05 is too high) - Use `--jinja` flag for llama.cpp
译文排队中 · 当前显示来源原文
Uncensored version of GPT-OSS 20B. This is the aggressive variant - tuned harder for fewer refusals.
译文排队中 · 当前显示来源原文
Uncensored version of GPT-OSS 20B. This is the balanced variant - more conservative tuning that prioritizes keeping the model's general capabilities intact. If you want fewer refusals and don't mind being a bit more experimental, check out the Aggressive variant.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. Uncensored version of GPT-OSS 120B by OpenAI. This is the aggressive variant - tuned harder for fewer refusals. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
译文排队中 · 当前显示来源原文
> **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. This release has NVIDIA's GenRM (generative reward model) fully removed. GenRM acts as an internal critic that scores and filters the model's own outputs — effectively a second layer of censorship on top of the base refusals. Removing it gives you the raw model output without any self-censoring.
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. **Non-thinking mode:** - Add `/no_think` at the end of your prompt, or - `temperature=0.7` - `top_p=0.8` - `top_k=20` - `min_p=0` **Important:** - Use `--jinja` flag for proper chat template handling - Thinking mode produces ` ... ` tags before responses
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring — model is fully unlocked and won't refuse prompts. Disclaimers that were present in previous releases have been significantly reduced in this version.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring with more thorough refusal removal. If this variant is too loose for your use case, a Balanced variant may follow.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring with more thorough refusal removal. If this variant is too loose for your use case, a Balanced variant may follow.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring — model is fully unlocked and won't refuse prompts. May occasionally append short disclaimers (baked into base model training, not refusals) but full content is always generated.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring with more thorough refusal removal. If this variant is too loose for your use case, a Balanced variant may follow.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals. Stronger uncensoring with more thorough refusal removal. If this variant is too loose for your use case, a Balanced variant may follow.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. > **Not sure which variant to pick?** 99.9%+ of users should use **Balanced** — same 0/465 refusal rate, more stable sampling, great for agentic coding / tool-use / reasoning / creative writing. Pick **Aggressive** only if you specifically want the model to skip its preamble on hardcore prompts. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended — just without the refusals.
译文排队中 · 当前显示来源原文
> **Join the Discord** for updates, roadmaps, projects, or just to chat. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads. No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
译文排队中 · 当前显示来源原文
> **HauhauCS FastMTP: up to 3.02x document TG and 1.93x reasoning TG versus non-MTP — plus up to 35.2% more document TG and 21.1% more reasoning TG than standard embedded MTP.** > **Join the Discord** for updates, roadmaps, projects, or just to chat. This is the **Aggressive variant**: direct answers, no refusal behavior, and minimal preamble on hard prompts.
译文排队中 · 当前显示来源原文
可以调整厂商、分类或输入输出条件；新模型正持续收录。
调整月度 Token 用量，直观比较不同模型的预估 API 开支。
使用已采集且单位可比的 USD / 百万 Token 渠道报价。报价为采集时快照；不计缓存优惠、阶梯价、推理额外输出与税费，请以来源的当前报价为准。
先区分输入、输出和缓存价格，再使用同一任务和用量估算总费用。图像、音频和视频可能按张、时长或分辨率计费，不能直接与 Token 单价比较。
不一定。必须逐一核对具体模型版本的许可证、商用限制及使用政策。自己部署还需要考虑显存、推理资源和运维费用。
榜单取决于测试任务、版本与评测方法。建议用实际业务样本比较准确性、时延、稳定性和成本，再选择模型。
