Open-weight model
Qwen3.8-27B-heretic-dflash-fp16-ov
by Blah Blah blaj/Qwen3.8-27B-heretic-dflash-fp16-ov
Qwen3.8-27B-heretic-dflash-fp16-ov is an open-weight model from Blah Blah, released under Apache License 2.0. It has 262,144-token context. Its published files total 3.5 GB.
The DFlash speculative-decoding draft for heretic-org/Qwen3.8-27B-heretic-ara, converted to OpenVINO IR at fp16 so it can be paired with an OpenVINO target in OVMS or OpenVINO GenAI. Not a standalone model.
Model Card
By Blah Blah, published under apache-2.0, revision f43ccf4a4708.
The DFlash speculative-decoding draft for heretic-org/Qwen3.8-27B-heretic-ara, converted to OpenVINO IR at fp16 so it can be paired with an OpenVINO target in OVMS or OpenVINO GenAI. Not a standalone model. It consumes the target's intermediate hidden states and proposes candidate tokens; the target verifies them. hiddenstates carries the 5 layers named by dflashconfig.targetlayerids (1,16,31,46,61), concatenated on the last axis. The target must export hiddenstatesdecoderlayers RT-info; the Optimum Intel export carries it automatically. OVMS logs Draft model strategy: DFlash when the pairing is accepted. Exported with Optimum Intel (text-generation-with-past, --trust-remote-code…
Read Blah Blah's full model card
Qwen3.8-27B-heretic-ara DFlash draft — OpenVINO IR (fp16)
The DFlash speculative-decoding draft for heretic-org/Qwen3.8-27B-heretic-ara,
converted to OpenVINO IR at fp16 so it can be paired with an OpenVINO target in OVMS
or OpenVINO GenAI.
Not a standalone model. It consumes the target's intermediate hidden states and proposes candidate tokens; the target verifies them.
I/O contract
inputs_embeds [?,?,5120] float32
hidden_states [?,?,25600] float32 # 5 target layers x 5120, concatenated
position_ids [?,?] int64
attention_mask [?,?] int64
beam_idx [?] int32
-> last_hidden_state [?,?,5120] float32
hidden_states carries the 5 layers named by dflash_config.target_layer_ids
(1,16,31,46,61), concatenated on the last axis. The target must export
hidden_states_decoder_layers RT-info; the Optimum Intel export carries it automatically.
Pairing (OVMS)
node_options: {
[type.googleapis.com / mediapipe.LLMCalculatorOptions]: {
models_path: "."
draft_models_path: "/path/to/this/draft"
}
}
OVMS logs Draft model strategy: DFlash when the pairing is accepted.
Conversion
Exported with Optimum Intel (text-generation-with-past, --trust-remote-code
--disable-convert-tokenizer), Transformers 5.2.0, OpenVINO 2026.3.1.
The draft declares architectures: ["DFlashDraftModel"], which Optimum Intel maps to its
built-in DFlash patcher. --trust-remote-code is required so that mapping is applied;
the config's own auto_map entry is not used.
Acknowledgements
Draft weights from jfan/Qwen3.8-27B-heretic-dflash, trained on
heretic-org/Qwen3.8-27B-heretic-ara hidden states. Apache-2.0.
Configuration
- Architecture
- DFlashDraftModel
- Context length (tokens)
- 262,144
- Layers
- 5
- Hidden size
- 5,120
- Feed-forward size
- 17,408
- Attention heads
- 32
- Key/value heads
- 8
- Head dimension
- 128
- Vocabulary size
- 248,320
- Sliding window (tokens)
- 2,048
- Model type
- qwen3
Identity and Version
- Repository
- blaj/Qwen3.8-27B-heretic-dflash-fp16-ov
- Publisher
- Blah Blah
- Task
- Not stated by the source
- Modality
- Other
- Library
- openvino
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- f43ccf4a4708cea2eeae4a3fb7abdcc81b1e89bc
- First published
- 2026-10-01
- Last updated
- 2026-10-01
Files and Weights
8 files, 3.5 GB in total. The weights are 1 file totalling 3.5 GB in bin.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| openvino_model.bin | Weights | 3.5 GB | 8c2862da8be8 |
| config.json | Configuration | 1.5 KB | — |
| generation_config.json | Configuration | 141 B | — |
| README.md | Documentation | 2.0 KB | — |
| openvino_model.xml | Other | 654.7 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
| tokenizer.json | Tokenizer | 1.6 KB | — |
| tokenizer_config.json | Tokenizer | 289 B | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 3.5 GB
Released by Blah Blah through its official repository on Hugging Face. Read the license.
Built From
- Derived from jfan/Qwen3.8-27B-heretic-dflash
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.5 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen3.8-27B-heretic-dflash-fp16-ov
Can I use Qwen3.8-27B-heretic-dflash-fp16-ov commercially?
Yes. Qwen3.8-27B-heretic-dflash-fp16-ov is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is Qwen3.8-27B-heretic-dflash-fp16-ov's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.