Qwen3.6-27B Fable-Fusion-711 Uncensored Heretic NEO-MAX MTPLX 8bit MLX
Specification
- Parameters
- 27B
- Architecture
- qwen3_5 (same fine-tune as M15, MLX repack by a different, unproven quantizer)
- Size on disk
- 28 GB
- Quantization
- 8bit
- Format
- MLX
- Runtime
- LM Studio
- Class
- Standard
- Reasoning (CoT)
- Yes — emits reasoning tokens
- Internal ID
- M19
- Mean speed
- 15.7 tok/s across suites
- Stall census
- 5 stalls in 59 observed tests (8.5%)
- Reasoning appetite
- 2,767 tokens mean · 16,277 max
- Model card
- huggingface.co
Suite results
General capability (13-task real-workload suite) 1a 248.3 / 260 19.1 13/13 n=1.1 16.5 run 2026-08-03
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| A1 | Dev/Ops Scripting | 14/20 | 17.5 | |
| A2 | Dev/Ops Scripting | 17/20 | 16.7 | |
| A3 | Dev/Ops Scripting | 20/20 | 16 | |
| A4 | Dev/Ops Scripting | 17/20 | 15.5 | |
| A5 | Dev/Ops Scripting | 20/20 | 16.3 | |
| A6 | Dev/Ops Scripting | 20/20 | 16.8 | |
| B1 | Document Processing | 20/20 | 16.7 | |
| B2 | Document Processing | 20/20 | 16.3 | |
| B3 | Document Processing | 20/20 | 15.7 | |
| C1 | Content Production | 20/20 | 16.4 | |
| C2 | Content Production | 20/20 | 16.6 | |
| C3 | Content Production | 20/20 | 17 | |
| C4 | Content Production | 20/20 | 17 |
Category mean Dev/Ops Scripting 18Document Processing 20Content Production 20
Agentic tool-calling & protocol adherence 1b 158.4 / 160 19.8 8/8 n=1 17.8 run 2026-08-04
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| WA1 | Tool Calling & Protocol Adherence | 20/20 | 15.9 | |
| WA2 | Tool Calling & Protocol Adherence | 20/20 | 16.4 | |
| WA3 | Tool Calling & Protocol Adherence | 20/20 | 16.7 | |
| WA4 | Tool Calling & Protocol Adherence | 20/20 | 15.9 | |
| WA5 | Tool Calling & Protocol Adherence | 20/20 | 17.9 | |
| WB1 | Agent Robustness & State Management | 20/20 | 17.7 | |
| WB2 | Agent Robustness & State Management | 20/20 | 17.1 | |
| WB3 | Agent Robustness & State Management | 18/20 | 16.9 |
Category mean Tool Calling & Protocol Adherence 20Agent Robustness & State Management 19.3
Coding depth 1c 159.2 / 180 19.9 8/9 ● n=1 16.8 run 2026-09-15
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| K1 | Coding — Concurrency | 20/20 | 16.8 | |
| K2 | Coding — Algorithms | 20/20 | 16.7 | |
| K3 | Coding — Security Review | 20/20 | 17 | |
| K4 | Coding — Performance | 20/20 | 16.9 | |
| K5 | Coding — Refactoring Under Constraints | 20/20 | 16.7 | |
| K6 | Coding — Test Writing | 20/20 | 17 | |
| K7 | Coding — React Component (typed) | 19/20 | 17.9 | |
| K8 | Coding — TypeScript Type System | 20/20 | 17.4 |
Category mean Coding — Concurrency 20Coding — Algorithms 20Coding — Security Review 20Coding — Performance 20Coding — Refactoring Under Constraints 20Coding — Test Writing 20Coding — React Component (typed) 19Coding — TypeScript Type System 20
Doc/OCR vision 1d 160 / 160 20 8/8 n=1 16.9 run 2026-08-06
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| D1 | Doc/OCR — Transcription | 20/20 | 15.1 | |
| D2 | Doc/OCR — Document QA | 20/20 | 13.5 | |
| D3 | Doc/OCR — Field Extraction | 20/20 | 14.2 | |
| D4 | Doc/OCR — Document QA | 20/20 | 15 | |
| D5 | Doc/OCR — Transcription | 20/20 | 14.8 | |
| D6 | Doc/OCR — Field Extraction | 20/20 | 14.8 | |
| D7 | Doc/OCR — Degraded Input | 20/20 | 16 | |
| D8 | Doc/OCR — Degraded Input | 20/20 | 15.4 |
Category mean Doc/OCR — Transcription 20Doc/OCR — Document QA 20Doc/OCR — Field Extraction 20Doc/OCR — Degraded Input 20
Doc/OCR — real-degraded tier 1d2 58 / 80 14.5 4/4 n=1 14.8 run 2026-08-14
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| D9 | Doc/OCR — Real Degraded Transcription | 15/20 | 14.4 | |
| D10 | Doc/OCR — Handwriting Extraction | 19/20 | 10.6 | |
| D11 | Doc/OCR — Degraded Table Transcription | 14/20 | 13.9 | |
| D12 | Doc/OCR — Degraded Table QA | 10/20 | 14.2 |
Category mean Doc/OCR — Real Degraded Transcription 15Doc/OCR — Handwriting Extraction 19Doc/OCR — Degraded Table Transcription 14Doc/OCR — Degraded Table QA 10
Content-production depth 1e 115.2 / 120 19.2 6/6 n=1 16.5 run 2026-08-04
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| E1 | SOW & Requirements Decomposition | 20/20 | 16.5 | |
| E2 | SOW & Requirements Decomposition | 20/20 | 16.4 | |
| E3 | Executive & Proposal Content | 18/20 | 16.4 | |
| E4 | Executive & Proposal Content | 20/20 | 16.7 | |
| E5 | Executive & Proposal Content | 17/20 | 16.7 | |
| E6 | Executive & Proposal Content | 20/20 | 16.1 |
Category mean SOW & Requirements Decomposition 20Executive & Proposal Content 18.8
Long-context multi-needle (MRCR) 1g2 44.1 / 60 14.7 3/3 n=1 13.8 run 2026-09-19
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| L7 | Long-Context — Multi-Needle Retrieval (MRCR) | 20/20 | 11.5 | |
| L8 | Long-Context — Multi-Needle Retrieval (MRCR) | 20/20 | 9 | |
| L9 | Long-Context — Multi-Needle Retrieval (MRCR) | 4/20 | 12.5 |
Production replay (real agent workload) 1i 127.2 / 240 10.6 12/12 n=1 12.2 run 2026-08-16
| Test | Category | Score | tok/s | |
|---|---|---|---|---|
| I1 | Nightly Fix Replay | 0/20 | 12.2 | |
| I2 | Nightly Fix Replay | 9/20 | 11.9 | |
| I3 | Nightly Fix Replay | 7/20 | 9 | |
| I4 | Nightly Fix Replay | 0/20 | 12.3 | |
| I5 | Nightly Fix Replay | 0/20 | 12.3 | |
| I6 | Nightly Fix Replay | 0/20 | 12.1 | |
| I7 | Audit Judgment | 15/20 | 13 | |
| I8 | Audit Judgment | 20/20 | 13 | |
| I9 | Cron Dependency Reasoning | 20/20 | 12.2 | |
| I10 | Cron Dependency Reasoning | 18/20 | 12.2 | |
| I11 | Curation Judgment | 20/20 | 12.9 | |
| I12 | Curation Judgment | 18/20 | 13 |
Category mean Nightly Fix Replay 2.7Audit Judgment 17.5Cron Dependency Reasoning 19Curation Judgment 19
Each test is LLM-judged 0–20 against a fixed rubric; suite max = tests × 20. An ADJ chip marks a cell carrying a subjective adjustment from the maintainer; the judged score is the one shown, is unchanged, and remains what the axes, ranks and tie bands are computed from. Hover for the reason. Rows with a ▶ expand to the per-test breakdown. Where fewer tests ran than the suite total, unrun tests count as zero toward the axis score. See methodology.