Maxwell-V1-Flash APEX-I V2.1 β€” MiniPlus

APEX Quantization Series by IsValorum

Support on Ko-fi

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.

Run it in V-Agent

Download V-Agent for Windows 10/11

Get the latest Windows installer Β· Release notes

Windows 10/11 Four local engines Seven MCP servers Apache 2.0 license

Your model. Your PC. Real tools. V-Agent is a free, open-source Windows desktop agent developed by the author of these APEX quantizations. Run GGUF locally, or connect a supported remote API.

  • Four native local engines: llama.cpp (CUDA, Vulkan, CPU), Strata / PolyStrata, Strata Rust / RustingStrata, and optional Project Maya. Use official builds where available or select compatible custom builds; Project Maya has its own prerequisites and setup.
  • Real agent workflows: seven integrated MCP servers for files and documents, commands, DuckDuckGo web search and page reading, autonomous browser, desktop tools, HTTP downloads and Hugging Face. Access modes, per-action approvals and isolated workspaces are user-controlled.
  • Find and run GGUF models: browse Hugging Face with hardware-fit estimates, import from LM Studio or Unsloth Studio, and download model files with SHA-256 verification. Local engine options adapt to the executable's reported capabilities.
  • Code checks and privacy: syntax and HTML runtime checks, context management and recovery from some transient failures. Local inference requires no V-Agent account or telemetry; optional web tools and remote APIs access their respective networks.

Open Models, search Maxwell-V1-Flash-APEX-I-MiniPlus-V2.1-GGUF, choose a compatible GGUF and confirm the download. V-Agent guides engine and model loading where supported.

Family cross-links

Tier Footprint Base Abliterated
MiniPlus 14.31 GiB Current β€”
NanoPlus 12.08 GiB Base β€”

Architecture and precision

  • Base model: maxwell-model/Maxwell-V1-Flash
  • Architecture: maxwell-moe
  • Quantization source precision: BF16
  • iMatrix: AtomicChat/Maxwell-V1-Flash-GGUF-metrics imatrix/imatrix.gguf, 4,766,918 tokens, SHA256 b6bbfdd99027e5c3dac9e97bafe9771719fd87e7fe90b2c979e116ef7e29ea69 (used file; external provenance not inferred)
  • Recurrent protections (if present): F32 (271 tensors observed)

Surgical Tensor Quantization Map (Audited from GGUF)

The table below is generated from the audited physical GGUF.

Complete per-tensor map.

Layer Group Sub-Component / Tensor Qty Precision Engineering Rationale
Global tensors output.weight 1 Q6_K Audited global tensor
Global tensors token_embd.weight 1 Q4_K Audited global tensor
Attention and gates blk.{0..44}.attn_k.weight 45 Q4_K Audited attention projection
Attention and gates blk.{0..44}.attn_output.weight 45 Q6_K Audited attention projection
Attention and gates blk.{0..44}.attn_q.weight 45 Q4_K Audited attention projection
Attention and gates blk.{0..44}.attn_v.weight 45 Q4_K Audited attention projection
Dense FFN / MLP blk.{0..2}.ffn_down.weight 3 IQ4_XS Audited dense MLP projection
Dense FFN / MLP blk.{0..2}.ffn_gate.weight 3 IQ3_S Audited dense MLP projection
Dense FFN / MLP blk.{0..2}.ffn_up.weight 3 IQ3_S Audited dense MLP projection
Shared experts blk.{3..44}.ffn_down_shexp.weight 42 Q5_K Audited shared-expert projection
Shared experts blk.{3..44}.ffn_gate_shexp.weight 42 Q5_K Audited shared-expert projection
Shared experts blk.{3..44}.ffn_up_shexp.weight 42 Q5_K Audited shared-expert projection
Routed experts blk.{3..12,35..44}.ffn_down_exps.weight 20 Q3_K Audited routed-expert projection
Routed experts blk.{13..34}.ffn_down_exps.weight 22 IQ3_XXS Audited routed-expert projection
Routed experts blk.{3..12,35..44}.ffn_gate_exps.weight 20 Q3_K Audited routed-expert projection
Routed experts blk.{13..34}.ffn_gate_exps.weight 22 IQ3_XXS Audited routed-expert projection
Routed experts blk.{3..12,35..44}.ffn_up_exps.weight 20 Q3_K Audited routed-expert projection
Routed experts blk.{13..34}.ffn_up_exps.weight 22 IQ3_XXS Audited routed-expert projection
Routing and indexer blk.{3..44}.ffn_gate_inp.weight 42 F32 Audited routing/indexer parameter
Routing and indexer blk.{0..44}.indexer.k_proj.weight 45 BF16 Audited routing/indexer parameter
Routing and indexer blk.{0..44}.indexer.q_proj.weight 45 BF16 Audited routing/indexer parameter
Hyperconnections blk.{0..44}.hc_attn_base.weight 45 F32 Audited hyperconnection parameter
Hyperconnections blk.{0..44}.hc_attn_fn.weight 45 F32 Audited hyperconnection parameter
Hyperconnections blk.{0..44}.hc_attn_scale.weight 45 F32 Audited hyperconnection parameter
Hyperconnections blk.{0..44}.hc_ffn_base.weight 45 F32 Audited hyperconnection parameter
Hyperconnections blk.{0..44}.hc_ffn_fn.weight 45 F32 Audited hyperconnection parameter
Hyperconnections blk.{0..44}.hc_ffn_scale.weight 45 F32 Audited hyperconnection parameter
Hyperconnections output_hc_base.weight 1 F32 Audited hyperconnection parameter
Hyperconnections output_hc_fn.weight 1 F32 Audited hyperconnection parameter
Hyperconnections output_hc_scale.weight 1 F32 Audited hyperconnection parameter
Normalizations blk.{0..44}.attn_k_norm.weight 45 F32 Verified physical normalization type
Normalizations blk.{0..44}.attn_norm.weight 45 F32 Verified physical normalization type
Normalizations blk.{0..44}.attn_q_norm.weight 45 F32 Verified physical normalization type
Normalizations blk.{0..44}.ffn_norm.weight 45 F32 Verified physical normalization type
Normalizations blk.{0..44}.indexer.k_norm.weight 45 F32 Verified physical normalization type
Normalizations blk.{0..44}.indexer.q_norm.weight 45 F32 Verified physical normalization type
Normalizations output_norm.weight 1 F32 Verified physical normalization type
Other block components blk.{3..44}.exp_probs_b.bias 42 F32 Audited structural parameter
Downloads last month
74
GGUF
Model size
36B params
Architecture
maxwell-moe
Hardware compatibility
Log In to add your hardware

We're not able to determine the quantization variants.

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for IsValorum/Maxwell-V1-Flash-APEX-I-MiniPlus-V2.1-GGUF

Quantized
(8)
this model

Collection including IsValorum/Maxwell-V1-Flash-APEX-I-MiniPlus-V2.1-GGUF