Qwopus3.6-27B-v2-MTPLX-4bit-Speed

Qwopus3.6-27B-v2 · MTPLX 4-bit Speed

The speed-focused MTPLX build of Qwopus3.6-27B-v2 — native multi-token-prediction speculative decoding on Apple Silicon, no external drafter, exact rejection sampling, so sampling behaves exactly like normal decoding, just faster. Higher-precision sibling: nom666/Qwopus3.6-27B-v2-MTPLX-8bit-Quality.

Forged with mtplx forge build from the original BF16 Jackrong/Qwopus3.6-27B-v2:

• Body: flat 4-bit MLX affine quantization, group size 64
• MTP head: preserved in BF16 (mtp_policy: keep_bf16), packed as mtp.safetensors sidecar
• Size: ~15 GB (runs comfortably on 24 GB Macs at moderate context; 32 GB+ for long context)
• Calibrated MTP contract included (mtplx_runtime.json)

Measured performance (Apple M5 Max, 128 GB, MTPLX 1.0.3)

Verification suite: long-code-uncapped, 2048-token budget. For reference, Qwopus-v2 Q6_K on llama.cpp with MTP (n=2) does ~24–26 tok/s on the same machine.

Usage

brew install youssofal/mtplx/mtplx   # or pipx install mtplx
mtplx pull nom666/Qwopus3.6-27B-v2-MTPLX-4bit-Speed
mtplx quickstart --model nom666/Qwopus3.6-27B-v2-MTPLX-4bit-Speed \
  --depth 3 --paged-kv-quantization q8 --batching-preset agent --reasoning off

Serves OpenAI-compatible (/v1/chat/completions) and Anthropic-compatible (/v1/messages) endpoints with warm-prefix KV reuse, SSD session cache, continuous batching, and vision support. Full 262144-token context; only 16 of 64 layers carry KV (hybrid Gated DeltaNet architecture), so KV at 256K is ~16 GiB BF16 / ~8 GiB q8.

Notes

• Runtime contract tier is forge-local: verified on the forging machine (M5 Max). MTPLX loads it with an honest provenance note.
• Prefer the 8-bit Quality sibling for maximum-fidelity agentic coding; this 4-bit build is the throughput pick.
• Quantized with MTPLX 1.0.3 Forge. All credit for the fine-tune to Jackrong; base model Qwen3.6-27B (Apache-2.0).