Qwopus 3.6 27B v2 MTP
Fine-tune of Qwen3.6 27B
composite -1.1compare to base
| Axis | Fine-tune | Base | Delta |
|---|---|---|---|
| Knowledge | 80.4 | 83.4 | -3.0 |
| Instruction | 57.8 | 65.0 | -7.2 |
| Coding | 29.8 | 25.5 | +4.3 |
| Math | 25.9 | 26.6 | -0.7 |
| Agentic | 9.4 | 8.3 | +1.1 |
Qwen3.6 27B
Every measured variant of this model: how much quality each quant keeps, and the VRAM and speed it costs to run.
VRAM footprint vs Local Intelligence Index
LB-2026-07.2 | 25/22.5/22.5/22.5/7.5. Where this model and current-lane family runs land vs the frontier anchors.
Swipe horizontally to inspect the full chart →
This modelFamily fine-tunesDashed vertical lines mark common VRAM tiers.Amber points are synthetic demo preview data.
Time to complete the benchmark
measured full-suite runs · wall time on each run's rig
Runs are ordered by elapsed time, shortest first. Output length affects totals — this is not an inference-speed ranking.
Elapsed time for this exact full-suite run; not a general model-speed measurement.
Variant profiles
Complete rows are ordered by Local Intelligence Index; partial rows show their measured axes but are not ranked. The VRAM/Fits columns (8Kcontext) tell you what your card needs. Ranks are within this family's variants.
Column guide
Variant — quant label plus, where sha-verified, the publisher of the exact weights file benchmarked; rows without a source line predate artifact identity records.
VRAM @8k — model weights, KV cache, and runtime headroom at 8k context.
Fits — the smallest common GPU VRAM tier above that estimate.
Prefill tok/s — prompt-processing speed.
Decode tok/s — generated-token speed after the prompt.
Overall tok/s — completion throughput across the full benchmark.
File size — the benchmarked model artifact on disk.
Runtime — the serving engine and version.
Run — the immutable benchmark receipt. Live rows link to the public submission record until the run is baked into the static site.
Pick the largest quant whose VRAM @8k fits your card.
Swipe horizontally for all variant metrics →
| Rank (this family) | Variant | Local Intelligence IndexLB-2026-07.2 | 25/22.5/22.5/22.5/7.5 | Agentic | Knowledge | Instruction | Coding | Math | VRAM @8k | Fits | Prefill tok/s | Decode tok/s | Overall tok/s | File size | Runtime | Run |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Q4_K_Mbestby lmstudio-community | 43.2±2.9 Agentic 8.3 / Knowledge 83.4 / Instruction 65.0 / Coding 25.5 / Math 26.6 | 8.3±7.4 | 83.4±3.9 | 65.0±5.4 | 25.5±7.1 | 26.6±7.2 | 19.5 GB | 24 GB | 3,112.1 | 74.4 | 69.4 | 17.1 GB | llama.cppb9852/fd1a05791 | receipt |
| 2 | fine-tuneQwopus 3.6 27B v2 MTPQ4_K_M | 42.1±3.0 Agentic 9.4 / Knowledge 80.4 / Instruction 57.8 / Coding 29.8 / Math 25.9 | 9.4±8.0 | 80.4±4.2 | 57.8±5.8 | 29.8±7.8 | 25.9±7.2 | 18.1 GB | n/a | 2,729.1 | 71 | 63.8 | 15.7 GB | llama.cppb9852/fd1a05791 | receipt |
| — | Q5_K_Mby unsloth | no run yet | — | — | — | — | — | 22.2 GB | 24 GB | — | 19.8 GB | — | benchmark it | ||
| — | UD-Q2_K_XLby unsloth | no run yet | — | — | — | — | — | 14.2 GB | n/a | — | 11.8 GB | — | benchmark it |
vs fine-tunes
Fine-tune comparison
Qwopus 3.6 27B v1 Preview
Fine-tune of Qwen3.6 27B
composite n/acompare to base
fine-tune not yet benchmarked
Measured axis deltas appear after both rows have board data.
Qwen3.6 27B MTP pi tune
Fine-tune of Qwen3.6 27B
composite n/acompare to base
fine-tune not yet benchmarked
Measured axis deltas appear after both rows have board data.
Qwen3.6 27B AEON Ultimate Uncensored BF16
Fine-tune of Qwen3.6 27B
composite n/acompare to base
fine-tune not yet benchmarked
Measured axis deltas appear after both rows have board data.
Bonsai 27B Ternary
Fine-tune of Qwen3.6 27B
composite n/acompare to base
fine-tune not yet benchmarked
Measured axis deltas appear after both rows have board data.