Quick Navigation Index
Independently computed using dedicated cloud compute clusters, funded 100% out-of-pocket.
If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies.
- Optimization History & Transparency Notice
- Quality Spectrum: APEX-I-NanoPlus vs. Standard Flat Quantizations
- Model Files & Technical Specifications
- Surgical Tensor Quantization Map (Audited from GGUF)
- Inference Quickstart
- CRITICAL: Coding Syntax & Repeat Penalty Advisory (Preventing Character Swapping)
- Hardened Agentic Chat Template & Reasoning Effort
- Optional Support
[!WARNING]
EXPERIMENTAL PRE-RELEASE NOTICE: ENGLISH-ONLY CODING SPECIALIST
This model suite is quantized from Jab1718/qwen3.8-flash-coder-85gb-bf16, which is an intermediate experimental slice created using
moe-slice(352 out of 512 routed experts were permanently pruned exclusively against English Python and SWE-bench calibration datasets).
- English Coding Only: This model is strictly designed for programming, code completion, refactoring, and agentic tool-calling in English.
- Severe Multilingual & General Degradation: Because conversational and multilingual experts were pruned and the upstream author has not yet released the recovery fine-tuning pass, this model severely degrades and outputs broken text in languages other than English (e.g., Spanish, French, German, etc.) or in general chit-chat.
- Incompatible with Strata Engine: This model uses a 160-expert layout with decoupled n-gram tables; it is not compatible with Strata Engine (which requires the 512-expert monolith and 51B PLE tables). Run using stock
llama.cpp(llama-server) or LM Studio.
Qwen3.8-Flash-Coder APEX-I-NanoPlus GGUF
The Next-Generation Frontier MoE · Extreme 18GB Footprint · Fast System RAM Streaming & Massive Context on 16GB–24GB VRAM
[!NOTE]
EXPLORE THE COMPLETE QWEN3.8 FLASH CODER LINEUP
These are complementary APEX-I releases, not alternate downloads of the same model:
- Qwen3.8-Flash-Coder APEX-I-MiniPlus-V2.1 — specialist agentic coding MoE tuned for Q5–Q6 quality (21.77 GB / 3.45 BPW).
- Qwen3.8-Flash-Coder APEX-I-NanoPlus — ultra-compact footprint achieving solid Q4 quality (18.34 GB / 3.44 BPW).
- Qwen3.8-Flash-Coder-85GB Lossless BF16 GGUF — uncompressed reference baseline (85.30 GB / 16.00 BPW).
[!TIP]
EMPIRICAL BENCHMARK & QUALITY COMPARISON
Quantization Specification File Size (Disk) Memory Footprint (RAM/VRAM) Average BPW WikiText-2 Perplexity Delta PPL vs BF16 (%) Quality Tier Equivalent Uncompressed BF16 Reference 85.30 GB(79.44 GiB)79.44 GiB16.00 BPW 30.0975 +/- 0.1200Baseline (0.00%) Lossless Reference Baseline APEX-I-MiniPlus V2.1 21.77 GB(20.27 GiB)20.27 GiB3.45 BPW 30.1495 +/- 1.0089+0.0520 (+0.17%)Q5_K_L / Q6_K Tier Boundary APEX-I-NanoPlus (CURRENT) 18.34 GB(17.08 GiB)17.08 GiB3.44 BPW 34.4199 +/- 1.1591+4.3224 (+14.36%)Solid Q4_K_M Tier Standard Flat Q3_K_S 20.41 GB 19.01 GiB 3.10 BPW aprox. 30.75 - 31.20 +0.65 a +1.10 (+2.9%) High syntax degradation Generic APEX Mini (IQ2_S) 17.73 GB 16.51 GiB 2.50 BPW aprox. 31.60 - 33.10+ +1.50 a +3.00+ (+7.5%) Severe reasoning breakdown
Model Files & Technical Specifications
| File Name | File Size | Memory Footprint | BPW | Description |
|---|---|---|---|---|
Qwen3.8-Flash-Coder-85GB.APEX-I-NanoPlus.gguf |
18.34 GB (17.08 GiB) |
17.08 GiB |
3.44 BPW | Ultra-compact agentic coding MoE achieving solid Q4 quality with massive context capability |
- Base Model: Jab1718/qwen3.8-flash-coder-85gb-bf16
- Parameters: 42.6B total (approx. 3.7B active per token: 10 routed experts + 1 shared expert + dense backbone; 4.9B with vocabulary embeddings)
- Architecture:
Qwen4ExpForCausalLM(48 hybrid layers, 160 MoE experts with 10 active) - Context Length: 262,144 tokens (native 256K)
Surgical Tensor Quantization Map (Audited from GGUF)
| Layer Group | Sub-Component / Tensor | Qty | Precision | Engineering Rationale |
|---|---|---|---|---|
| Global Output Head | output.weight |
1 | Q6_K |
Armored in high-precision Q6_K to preserve token classification. |
| Global Embeddings | token_embd.weight |
1 | Q3_K |
Compact embedding representation across 248k vocabulary. |
| Attention & Highway Norms | attn_{q,k}_norm, hc_{attn,ffn}_norm |
120 | F32 |
100% uncompressed numerical stability across all 48 layers. |
| Expert Routers | blk.*.ffn_gate_inp, ffn_gate_inp_shexp |
96 | F32 |
100% uncompressed routing fidelity across 160 experts. |
| Attention Gates | blk.*.attn_gate.weight (36 SSM Layers) |
36 | Q8_0 |
High-precision attention gating across hybrid DeltaNet recurrence layers. |
| Linear Attention Projections | blk.*.attn_qkv.weight (36 SSM Layers) |
36 | Q3_K |
Efficient 3-bit quantization for SSM attention state inputs. |
| SSM Linear Output | blk.*.ssm_out.weight (36 SSM Layers) |
36 | Q5_K |
5-bit precision for linear state-space recurrence output. |
| Recurrent SSM Parameters | blk.*.ssm_{a,alpha,beta,conv1d,dt,norm} |
216 | F32 |
Guarded in uncompressed FP32 to prevent DeltaNet recurrent state drift. |
| Periodic Sparse Attention | blk.{3,7,...}.attn_{q,k,v}.weight (12 Layers) |
36 | Q4_K |
Quadratic attention checkpoints for deep retrieval. |
| Periodic Sparse Attention Output | blk.{3,7,...}.attn_output.weight (12 Layers) |
12 | Q6_K |
Armored attention output projection over deep context. |
| QSA Sparse Attention Indexer | blk.{3,7,...}.indexer.{q,k}_proj.weight |
24 | BF16 |
High-fidelity sparse indexer projections for query-stream attention routing. |
| QSA Indexer Norms | blk.{3,7,...}.indexer.{q,k}_norm.weight |
24 | F32 |
Uncompressed indexer layer normalizations. |
| Shared Foundation Experts | blk.*.ffn_down_shexp.weight (All 48 Layers) |
48 | Q5_0 |
ne0=640 in standard block-32; zero divisibility crashes. |
| Shared Foundation Experts | blk.*.ffn_{gate,up}_shexp.weight (All 48 Layers) |
96 | Q5_K |
Preserves core coding knowledge active on 100% of tokens. |
| Highway Connections (Down/Inject) | blk.*.hc_{attn,ffn}_{down,inject}.weight |
192 | Q6_K |
High-fidelity residual highway bypass. |
| Highway Connections (Up) | blk.*.hc_{attn,ffn}_up.weight |
96 | Q5_0 |
ne0=320 in standard block-32 format. |
| Routed MoE Down-Projections | blk.*.ffn_down_exps.weight (All 48 Layers) |
48 | IQ4_NL (36) / Q4_0 (12) |
Non-linear codebook quantization for SSM layers, Q4_0 for anchor layers. |
| Routed MoE Gate/Up Projections | blk.*.ffn_{gate,up}_exps.weight (All 48 Layers) |
96 | IQ2_S (36) / IQ2_XXS (36) / IQ3_XXS (24) |
Scaled expert density: sub-2.5 BPW on deep layers, IQ3_XXS on anchor layers. |
| Residual Output Highway | output_hc_{down,up}.weight |
2 | Q4_0 |
Low-rank residual highway projections at model termination. |
| Residual Output Highway Norm | output_hc_norm.weight |
1 | F32 |
Final residual normalization anchor. |
Inference Quickstart
llama-server \
-m Qwen3.8-Flash-Coder-85GB.APEX-I-NanoPlus.gguf \
--jinja \
-ngl 99 \
--ctx-size 65536 \
--temp 0.6 --top-p 0.95 --top-k 20 --min-p 0.00 \
--port 8080
[!IMPORTANT]
CRITICAL ADVISORY FOR CODING WORKFLOWS: PREVENTING SYNTAX & TOKEN SWAPPING
Disable repeat penalties (
repeat_penalty: 1.0,presence_penalty: 0.0,frequency_penalty: 0.0) and use--jinjato avoid syntax bracket substitutions.
7. Community Compute Fund & Student Support
All IsValorum quantizations will always remain completely free and open to the public without paywalls.
However, cloud GPU compute is expensive and funded entirely out-of-pocket. If you find these handcrafted builds to be of high quality and they benefit your day-to-day workflow, you can support future Day-0 releases on Ko-fi. Every contribution directly funds the cloud compute hours required for new model runs and helps support my university education:
(When supporting on Ko-fi, feel free to leave a note with your Hugging Face handle and the specific model you would like prioritized).