HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP (26B MoE, 4B active, Google QAT checkpoint) with a 1,048,576-token context baked in (4x the native 262,144), shipping with its MTP speculative-decoding draft head and vision tower. All numbers below were measured on these exact files.
Capability
Status
1M context
~91% mean recall across two seed sets (honesty note below)
MTP speculative decoding
249.1 to 369.2 tok/s (+48%), acceptance 0.679 (measured on this trunk, RTX 5090)
Vision
Verified July 6, 2026: reads image text and identifies objects
Uncensored
HauhauCS Balanced abliteration; trunk weights bit-identical to the source release
Needle-in-a-haystack
Full transparency: across two complete seed sets this trunk averages ~91 percent needle recall, dropping roughly one needle per rung at random depths, including inside the native 262K range. The official censored trunk scored 10/10 at 393K under the same harness, so this is a small flat abliteration tax, not a context-length failure. If a rare retrieval miss is unacceptable for your workload, use the 12B, which is certified clean. Every run, including the imperfect ones, is in results.jsonl. Rungs above 524K exceed 32 GB VRAM and have not been run yet; they are queued for larger hardware and will publish here as measured, pass or fail.
MTP speculative decoding
The draft head predicts ahead and the trunk verifies every token, so output is identical to standard decoding, only faster. Measured speedup on this uncensored trunk beats the ~35 percent claimed upstream.
Files
File
Size
Role
gemma4-26b-a4b-uncensored-1M-Q4_K_M.gguf
16.8 GB
Trunk, 1M baked, QAT 4-bit
mtp-gemma-4-26B-A4B-it.gguf
252 MB
MTP draft head, pair with -md
mmproj-gemma26b-hauhau.gguf
1.2 GB
Vision tower, pair with --mmproj
niah_heatmap.png, mtp_speedup.png, results.jsonl
small
Verification evidence
Every file, every mirror
Nothing was discontinued: every quant is one click away. Hugging Face carries the curated picks, ModelScope always carries everything, and Ollama serves ready-to-run tags.
On Ollama every tag ships with the vision tower bundled and the 1M rope metadata baked in.
Ollama (1M and vision work; Ollama has no speculative decoding yet, so the MTP head adds no speed there):
FROM ./gemma4-26b-a4b-uncensored-1M-Q4_K_M.gguf
RENDERER gemma4
PARSER gemma4
PARAMETER num_ctx 262144
The RENDERER and PARSER lines avoid imported-GGUF template bugs under tool-heavy use. Raise num_ctx as memory allows.
How this was built
YaRN rope-scaling metadata (factor 4.0 over native 262,144) baked into the GGUF header with gguf-py; weights are bit-identical to the HauhauCS release, no fine-tuning. Gemma 4's dual-rope design takes YaRN on its global-attention layers. Certification harness: 10 needles per rung at depths 5 to 95 percent, temperature 0, seeded prompts, f16 KV only. Method and tooling: github.com/satindergrewal/aviary-1m.
For base capability benchmarks see Google's official Gemma 4 cards; uncensoring quality versus the official trunk has not been independently benchmarked here.
How to actually use a 1M-context model
Habits that measurably help, from our RULER, hop and adherence testing across this fleet:
Re-state standing instructions near the end of long prompts; recency beats depth.
One big reference dump beats a long accumulated conversation. Fresh session per task.
After any compaction or summarization, repeat your active rules yourself.
Prefill at 500K+ takes real time on any hardware; stage your questions accordingly.
Know your quant: the results tables on this card show what each quant actually holds at depth; pick the strongest one your memory allows.
Credits
Base model and QAT: Google (Gemma license; its terms flow down to these files). Uncensoring and packaging: HauhauCS. MTP head: Unsloth (via the HauhauCS repo). 1M YaRN extension, benchmarking, and certification: SatGeze.
