Compare model versions

Diff

Before

<div style="display: flex; gap: 8px; align-items: center; margin-top: 10px; margin-bottom: 10px;">
  <a href="https://atomic.chat/?utm_source=huggingface&utm_medium=referral&utm_campaign=hf_ling_3_0_flash&utm_content=btn_atomic"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_atomic.png" width="162" alt="Atomic Chat"></a>
  <a href="https://discord.gg/8wGSsvmg4V"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_discord.png" width="119" alt="Discord"></a>
  <a href="https://github.com/AtomicBot-ai/Atomic-Chat"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_github.png" width="115" alt="GitHub"></a>
</div>
<ul style="margin: 0 0 12px 0;">
  <li>Ling 3.0 flash is a 124B MoE with 5.1B active parameters; its GGUF chat template includes thinking mode and tool calling.</li>
  <li>Our AD layouts have 31–41% lower mean KL than stock llama.cpp presets at comparable sizes. The <a href="https://huggingface.co/datasets/AtomicChat/Ling-3.0-flash-GGUF-metrics">raw measurements</a> and controls are public.</li>
  <li>From the team behind <a href="https://atomic.chat/?utm_source=huggingface&utm_medium=referral&utm_campaign=hf_ling_3_0_flash&utm_content=bullet_app">Atomic Chat</a>, the local AI app for chat and agents. These Ling files require our TurboQuant build; use the <a href="#run-it">CLI instructions below</a>. The tests documented below cover the CLI workflow, not app compatibility.</li>
</ul>
<hr style="margin: 0 0 16px 0;">

Quantizations of [`inclusionAI/Ling-3.0-flash`](https://huggingface.co/inclusionAI/Ling-3.0-flash):
124B total, 5.1B active, hybrid linear attention (35 KDA blocks interleaved 5:1 with 7 gated MLA
blocks) over a 512-expert MoE.

Bits are placed by hand rather than by the default rules, and the controls that prove it is worth
something are published next to the files. At the same size, our layout sits 31 to 41 % closer to
BF16 than what `llama-quantize` produces on its own

> **These files need a TurboQuant build.** `bailingmoe3` is in upstream llama.cpp, but we have some important bugfixes to it.
> Nothing has to be compiled, see [Run it](#run-it).

After

<div style="display: flex; gap: 8px; align-items: center; margin-top: 10px; margin-bottom: 10px;">
  <a href="https://atomic.chat/?utm_source=huggingface&utm_medium=referral&utm_campaign=hf_ling_3_0_flash&utm_content=btn_atomic"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_atomic.png" width="162" alt="Atomic Chat"></a>
  <a href="https://discord.gg/8wGSsvmg4V"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_discord.png" width="119" alt="Discord"></a>
  <a href="https://github.com/AtomicBot-ai/Atomic-Chat"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_github.png" width="115" alt="GitHub"></a>
  <a href="https://x.com/atomic_chat_hq"><img src="https://huggingface.co/AtomicChat/Qwen3.8-27B-GGUF/resolve/main/btn_x.png" width="108" alt="X"></a>
</div>
<ul style="margin: 0 0 12px 0;">
  <li>Ling 3.0 flash is a 124B MoE with 5.1B active parameters; its GGUF chat template includes thinking mode and tool calling.</li>
  <li>Our AD layouts have 31–41% lower mean KL than stock llama.cpp presets at comparable sizes. The <a href="https://huggingface.co/datasets/AtomicChat/Ling-3.0-flash-GGUF-metrics">raw measurements</a> and controls are public.</li>
  <li>From the team behind <a href="https://atomic.chat/?utm_source=huggingface&utm_medium=referral&utm_campaign=hf_ling_3_0_flash&utm_content=bullet_app">Atomic Chat</a>, the local AI app for chat and agents. These Ling files require our TurboQuant build; use the <a href="#run-it">CLI instructions below</a>. The tests documented below cover the CLI workflow, not app compatibility.</li>
</ul>
<hr style="margin: 0 0 16px 0;">

Quantizations of [`inclusionAI/Ling-3.0-flash`](https://huggingface.co/inclusionAI/Ling-3.0-flash):
124B total, 5.1B active, hybrid linear attention (35 KDA blocks interleaved 5:1 with 7 gated MLA
blocks) over a 512-expert MoE.

Bits are placed by hand rather than by the default rules, and the controls that prove it is worth
something are published next to the files. At the same size, our layout sits 31 to 41 % closer to
BF16 than what `llama-quantize` produces on its own

> **These files need a TurboQuant build.** `bailingmoe3` is in upstream llama.cpp, but we have some important bugfixes to it.
> Nothing has to be compiled, see [Run it](#run-it).