---
title: Qwen3.8-Flash-Next-E384-FP4-V2
canonical_url: "https://www.modelscope.ai/models/AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2"
md_url: "https://www.modelscope.ai/models/AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2.md"
repository: AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2
last_updated: 2026-10-27
license: other
pipeline_tag: image-text-to-text
tasks:
  - image-text-to-text
model_type:
  - qwen4_exp
architectures:
  - Qwen4ExpForConditionalGeneration
base_model:
  - autotrust/Qwen3.8-Flash-Next-E384-FP4-Local
base_model_relation: merge
parameters: 80.1B
tensor_type:
  - I64
  - U8
  - U32
  - F32
  - F8_E4M3
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - zh
downloads: 7
stars: 0
tags:
  - qwen3.8
  - moe
  - nvfp4
  - modelopt
  - mtp
  - rl
  - grpo
---

# Qwen3.8-Flash-Next-E384-FP4-V2

> Qwen3.8-Flash-Next-E384-FP4-V2 - An open-source model by AutoTrust-Labs on ModelScope. Qwen3.8-Flash-Next-E384-FP4-V2

AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2 is a 80.1B-parameter image-text-to-text model on ModelScope. licensed under other. derived from autotrust/Qwen3.8-Flash-Next-E384-FP4-Local.

- **Repository**: AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2
- **License**: other
- **Tasks**: image-text-to-text
- **Parameters**: 80.1B
- **Base model**: autotrust/Qwen3.8-Flash-Next-E384-FP4-Local
- **Tags**: qwen3.8, moe, nvfp4, modelopt, mtp, rl, grpo
- **Downloads**: 7
- **Stars**: 0
- **Last updated**: 2026-10-27

Source: https://www.modelscope.ai/models/AutoTrust-Labs/Qwen3.8-Flash-Next-E384-FP4-V2

---

# Qwen3.8-Flash-Next-E384-FP4-V2

Full merged checkpoint of the E384 FP4 model after GRPO-style reinforcement learning with a LoRA
adapter. **This is a complete model, not an adapter**: all 21 weight shards, the tokenizer, configs and
MTP files are included. It is an unofficial derivative and is not endorsed by Alibaba or NVIDIA.

## Serving

Validated with vLLM 0.31 on one B200. The included `vllm_patch/sitecustomize.py` is needed for this
checkpoint's FP4/CSF/PLE weight layout. From this directory:

```bash
export PYTHONPATH="$PWD/vllm_patch${PYTHONPATH:+:$PYTHONPATH}"
vllm serve . --served-model-name e384-rl --port 8000 \
  --tensor-parallel-size 1 --dtype bfloat16 --kv-cache-dtype fp8 \
  --quantization modelopt_mixed --mamba-cache-mode align \
  --enable-prefix-caching --enable-chunked-prefill \
  --max-model-len 72000 --max-num-seqs 128 --gpu-memory-utilization 0.90 \
  --reasoning-parser qwen3 --tool-call-parser qwen3_xml --enable-auto-tool-choice
```

`serve_vllm.sh` is the equivalent portable launcher. The model was benchmarked through this serving
path, including the FP8 KV cache and the Qwen reasoning/tool parsers.

## Training and checkpoint contents

- Three GRPO-style LoRA updates with verifiable rewards; the final round targets short-budget CVSS
  completion and structured findings emission.
- LoRA rank 32, alpha 64. The routed experts, router, shared expert and PLE are frozen.
- The GDN projection deltas are folded at full precision and those projections are stored as BF16;
  the corresponding entries are removed from `quantization_config.quantized_layers`. Other model
  tensors retain the base checkpoint format.
- The final adapter was merged into the weights before this repository was created. No adapter loading
  step is needed.

## Full benchmark results

Same benchmark harness and serving settings for all rows. These are single runs; GPQA has only 198
questions and is correspondingly noisy. AIME 2026 uses 30 problems x 4 samples with xhigh reasoning and
a 64k generation budget. Base numbers are the previously published reference numbers, not a rerun.

| benchmark | base reference | this model |
|---|---:|---:|
| GPQA-Diamond (low, 32k) | 77.3 | **75.76** |
| MMLU (14,042) | 86.9 | **87.37** |
| C-Eval val (1,346) | 89.0 | **89.97** |
| AIME 2026 (xhigh, 64k) | 93.3 | **95.0** |
| HumanEval (164, greedy) | 97.6 | **98.78** |
| MMMU val (900) | 79.4 | **78.22** |
| tool calling (800) | — | **77.88** |

Detailed per-task counts and truncation statistics are in `benchmark_results.json`.

### AIME reasoning length

Compared with the preceding RL checkpoint on the same 120 AIME samples, total completion tokens fell
2.05% (19,897 to 19,490 per answer on average), while accuracy rose from 92.5 to 95.0. The median was
essentially unchanged (16,528 to 16,675 tokens); the reduction came mainly from the long tail (P90
58,272 to 45,339) and fewer 65,536-token truncations (9 to 6). `completion_tokens` includes both the
reasoning trace and the final answer.

## Behavior notes and limitations

On the same training-prompt evaluation used during the final RL round, CVSS3 completion within a 2,200
token budget rose from 39% to 70%, CVSS4 from 18% to 60%, and structured-findings emission reward from
0.609 to 0.917. Math and guard rewards did not degrade in that evaluation. These are training-prompt
measurements, not an independent behavior benchmark.

**CVSS arithmetic is still not reliable**: among completed answers, exact agreement with the reference
implementation was about 4.5% for CVSS3 and 4.2% for CVSS4 in that evaluation. This checkpoint should not
be described as having fully fixed CVSS scoring. See `behavior_results.json` for the detailed comparison.