Gemma 4 26B-A4B 98e v7-coderx — code-maximal prune

20.8B params · 98 experts (30 dropped) · ~4B active · code-maximal drop map

A research checkpoint that takes Gemma-4-26B-A4B-it and drops ³⁰⁄₁₂₈ experts per layer using a code-maximal recipe on the rebuilt v7 competence maps (audited producers, 10 classes) — generic-code 3× + LiveCodeBench-medium 2× on a [24,40] per-layer floor, with no science or multilingual targeting. Same router, attention, and norms as base, plus the mandatory shared-FFN α=1.2 upweight every coder variant carries.

The strongest coder in the cohort: it spends its whole prune budget on code and lands LiveCodeBench-medium-55 at 98.18% and LCB-100 at 99.0% — the highest of any Gemma-4 prune to date, +1.8pp / +2.0pp past the unpruned 128e (96.36 / 97.0). The trade is graduate science (GPQA 48.48). If you need the science back without giving up the code profile, use the sibling v7-coder (GPQA 70.71, LCB-55 96.36).

Full model card & methodology: ManniX-ITA/gemma-4-A4B-98e-v7-coderx-it on Hugging Face.

Other formats: - GGUF (29 tiers, imatrix, CD-* per-layer mixes + F16 + mmproj): ManniX-ITA/gemma-4-A4B-98e-v7-coderx-it-GGUF - NVFP4A16 (native vLLM, ~13 GB): ManniX-ITA/gemma-4-A4B-98e-v7-coderx-NVFP4A16

Scores (Q6_K, llama.cpp, greedy, same host)

Reference columns on the same Q6_K run: unpruned 128e LCB-55 96.36 / LCB-100 97.00 / MultiPL-E 90.00; v6-coder LCB-55 92.73 / LCB-100 94.00. v7-coderx tops the cohort on every code/instruction axis; the budget is paid almost entirely on graduate science (GPQA 48.48, vs 128e 67.17).

Quantizations — HE+ / MultiPL-E-100 score, size & answer length

Every K-quant and CD tier was scored on HumanEval+ (164) and MultiPL-E-100 (llama.cpp, greedy T=0), with per-problem completion length from token_stats. bpw is the true bits-per-weight (8 × bytes ÷ 19,877,953,946). ⭐ marks a recommended pick.

Recommended picks:

• Q4_K_M ⭐ (13.24 GB) — best K-quant — 93.29% HE+ (13.2 GB).
• CD-Q3_K_L ⭐ (10.22 GB) — best overall — highest HE+ of any tier (93.90%) at 10.2 GB.
• CD-Q2_K ⭐ (8.82 GB) — smallest tier still in the 90%+ band — 90.24% HE+ at 8.8 GB.

The K-quant and CD tiers hold HE+ in the 90–93% band with length essentially identical to Q6_K; the 2-bit Q2_K_L / IQ2_XS tiers are the cliff (HE+ into the 80s/70s, token p90 blows out). The K-quant CD tiers are the recommended low-bit path — CD-IQ* i-quant bodies are not offered (the pruned MoE degenerates on an IQ-family body, score → 0). Prefer Q4_K_M or higher for production.

Head-to-head by file size — v7-coderx vs Qwen2.5-Coder-14B (iso-disk)

Pairing by tier name is misleading — this is a ~20.8B-total MoE and Qwen2.5-Coder-14B is a 14.7B dense model, so the same tier name lands at a different size. The fair comparison is iso-disk: at a given GB budget, which scores higher on HumanEval+? Same rig (RTX 3090, llama.cpp, greedy). Qwen GGUFs are bartowski’s (83–85% across the ladder). At every band the MoE runs lower bpw at the same disk and still scores higher.

Pull

ollama pull mannix/gemma4-98e-v7-coderx                 # :latest = Q4_K_M (best K-quant, 93.29% HE+)
ollama pull mannix/gemma4-98e-v7-coderx:CD-Q3_K_L       # ⭐ best overall — 93.90% HE+, 10.2 GB
ollama pull mannix/gemma4-98e-v7-coderx:Q6_K            # max fidelity (bench tier)
ollama pull mannix/gemma4-98e-v7-coderx:CD-Q2_K         # smallest 90%+ — 90.24% HE+, 8.8 GB
ollama pull mannix/gemma4-98e-v7-coderx:vision-Q4_K_M   # + SigLIP vision tower

Inherits Gemma 4’s thinking format — serve with the reasoning parser enabled (--reasoning-format deepseek --reasoning-budget 8192 on llama-server).

Derivative of Gemma 4 — Gemma Terms of Use.