README.md

---
license: apache-2.0
library_name: mlx
base_model: Zyphra/ZAYA1-8B
base_model_relation: quantized
pipeline_tag: text-generation
tags:
  - zaya
  - mixture-of-experts
  - hybrid-attention
  - cca-attention
  - mlx
  - apple-silicon
  - reasoning
  - tool-use
  - quantized
  - jang
  - jangtq
  - mxtq
  - jangtq-prestack
  - osaurus
quantization_config:
  family: jangtq
  profile: JANGTQ2
  group_size: 32
  expert_layout: split_switch_mlp
---

<p align="center"><img src="osaurus-x-banner.png" width="100%" alt="OsaurusAI"/></p>

# ZAYA1-8B-JANGTQ2

Quantized **Zyphra/ZAYA1-8B** for Apple Silicon runtimes.

| | |
|---|---|
| Source | [Zyphra/ZAYA1-8B](https://huggingface.co/Zyphra/ZAYA1-8B) |
| License | Apache-2.0, inherited from upstream |
| Format | JANGTQ2 |
| Modality | text |
| Bundle size | 2.77 GiB |
| Tensor keys | 1965 |
| Expert layout | Pre-stacked `zaya_block.experts.switch_mlp` |
| Runtime status | Generation coherence: NOT INDEPENDENTLY PASSED for the quantized runtime bundle (missing coherence report); published as a format/runtime bundle pending downstream ZAYA runtime validation. |

## Important Runtime Note

This bundle requires a ZAYA-aware JANGTQ runtime that implements CCA attention state plus pre-stacked `switch_mlp` TurboQuant experts.

ZAYA is not a stock `mlx_lm` architecture. It alternates CCA attention layers and top-1 MoE layers. Use this bundle only with a runtime that implements the ZAYA CCA state contract and the converted pre-stacked expert layout.

## Runtime Pin Required

Use a `vmlx-swift-lm` build that includes the ZAYA Swift runtime (`Libraries/MLXLLM/Models/Zaya.swift` + `MLXLMCommon/Cache/ZayaCCACache.swift` + `BatchEngine/BatchZayaCCACache.swift`). The first verified pin is commit `b9da180` or newer.


## Architecture Summary

- 80 decoder layers: alternating CCA attention and top-1 MoE
- Hidden size 2048, 16 query heads, 2 KV heads, head dim ?
- CCA state per attention layer: standard KV plus `conv_state [B,1280,2]` and `prev_hs [B,2048]`
- 16 routed experts per MoE layer, top-1 routing with MOD skip route
- Context length 131072, `rope_theta=5000000`

## Quantization

2-bit MXTQ routed experts + 8-bit affine non-routed tensors.

Passthrough floor for first release prep:

- `conv_qk.*`, `temp`, norms, residual scaling, router path, biases, and balancing biases are preserved as float tensors.
- Embeddings and `lm_head` use 8-bit affine in the prepared bundles.
- Text-only ZAYA1-8B has no vision_tower or LoRA tensors.
- `jangtq_runtime.safetensors` is included: true.

`mxtq_bits`:

```json
{
  "routed_expert": 2,
  "attention": 8,
  "router": 16,
  "embed_tokens": 8,
  "lm_head": 8,
  "cca_conv": 16,
  "norms_residual": 16
}
```

## Bundle Verification

- Safetensor headers scanned.
- Source tensor coverage checked.
- Converted bundles checked for `local_experts` removal.
- Converted expert tensors checked for pre-stacked `switch_mlp` layout.
- JANGTQ sidecars checked for the Swift runtime contract.
- Capabilities verified: family=zaya, supports_thinking=False, tool_parser=zaya_xml.
- Runtime coherence status recorded above.

## Runtime Smoke Tests

Before production use, run short deterministic prompts through the exact target runtime:

- `What is 2+2? Answer with only the number.`
- `What is the capital of France? Answer with one word.`
- One chat-template prompt with thinking disabled.
- One chat-template prompt with thinking enabled and enough output budget for the final answer.

The first public bundle release records bundle integrity and runtime contract checks. Full generation quality depends on a ZAYA-aware runtime implementation.

## Korean Summary

이 번들은 Zyphra/ZAYA1-8B를 Apple Silicon MLX/JANG 런타임용으로 양자화한 모델입니다. ZAYA의 CCA attention 상태와 MoE 라우팅을 정확히 구현한 런타임에서만 사용해야 합니다.

## Files

- `config.json` carries `weight_format=mxtq`, `zaya_expert_layout=split_switch_mlp`.
- `jang_config.json` carries `cache_subtype=zaya_cca`.
- Tokenizer files and chat template are preserved from the upstream source snapshot.