Following the widespread community adoption of the Qwen3.5 and Qwen3.6 series, we are pleased to introduce Qwen3.8, the most capable generation in the Qwen open-model family to date.
gemma-4-26B-A4B-it
google·Instruct·released 11 Mar 2026
ollama run gemma4:26bGemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants.
How much VRAM does gemma-4-26B-A4B-it need?
At Q4_K_M gemma-4-26B-A4B-it needs about 18 GB with an 8K context (18 GB at Q4, 53 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈18 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 8K | 16K | 32K | 64K | 128K | 256K |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈11 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈15 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈18 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈20 GB |
| Q6_KNear-losslessPractically lossless | ≈23 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈29 GB |
| FP16 / BF16LosslessFull precision, as released | ≈52 GB |
Estimates for 8K tokens of context: weights + KV cache + runtime overhead. All experts stay in memory, but only the active ones are read per token. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.
How to pick a quant
Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.
- LosslessF16, Q8_0
- Near-losslessQ6_K, Q5_K_M
- Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
- Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
- Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K
Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B
How much context fits?
At 8K context, Q4_K_M needs ≈18 GB(15 GB weights + 1.9 GB KV cache + 1.2 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Can the Tesla P40 run gemma-4-26B-A4B-it? →
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of gemma-4-26B-A4B-it
Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →
No medians yet. Be one of the first 3 to report a setup.
All reports
Run gemma-4-26B-A4B-it locally
--quantQ4_K_Mdefault; pick your GPU to tailor itunsloth/gemma-4-26B-A4B-it-GGUF
ollama run gemma4:26bThe library build picks its own default quantisation.
ollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_M llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_MDownloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.
lms get unsloth/gemma-4-26B-A4B-it-GGUF@UD-Q4_K_MThen load it from the LM Studio app or with `lms load`.
vllm serve google/gemma-4-26B-A4B-it vllm serve RedHatAI/gemma-4-26B-A4B-it-FP8-dynamicservices:
vllm:
image: vllm/vllm-openai:latest
command: google/gemma-4-26B-A4B-it
ports:
- "8000:8000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 8000.
sglang serve google/gemma-4-26B-A4B-it --host 0.0.0.0 --port 30000services:
sglang:
image: lmsysorg/sglang:latest
command: sglang serve google/gemma-4-26B-A4B-it --host 0.0.0.0 --port 30000
ports:
- "30000:30000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 30000.
from transformers import pipeline
pipe = pipeline("image-text-to-text", model="google/gemma-4-26B-A4B-it", dtype="auto", device_map="auto")
messages = [
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
{"type": "text", "text": "Describe this image."},
],
}
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| MMLU-Pro | 82.6% | self-reported |
| GPQA Diamond | 82.3% | self-reported |
| Humanity's Last Exam | 17.2% | self-reported |
| MMMU-Pro | 73.8% | self-reported |
| ParseBench | 58.5% | independent |
Self-reported scores come from the model card and are not independently verified. Rank is among all models with a score on that benchmark.
Quantised versions and downloads
| Repository | Format | Quant | Size |
|---|---|---|---|
| nvidia/Gemma-4-26B-A4B-NVFP4 | Safetensors | NVFP4 | — |
| RedHatAI/gemma-4-26B-A4B-it-NVFP4 | Safetensors | NVFP4 | — |
| RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic | Safetensors | FP8 | — |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-IQ2_XXS | 9.9 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-IQ2_M | 10 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-Q2_K_XL | 11 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-IQ3_S | 11 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-IQ3_XXS | 11 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-Q3_K_M | 13 GB |
| unsloth/gemma-4-26B-A4B-it-GGUF | GGUF | UD-Q3_K_XL | 13 GB |
About gemma-4-26B-A4B-it
Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages. Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: E2B, E4B, 12B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from high-end phones to laptops and servers, democratizing access to state-of-the-art AI. Gemma 4 introduces key capability and architectural advancements: • Reasoning – All models in the family are designed as highly capable reasoners, with configurable thinking modes. • Extended Multimodalities – Processes Text, Image with variable aspect ratio and resolution support (all models), Video, and Audio (featured natively on the E2B, E4B, and 12B models). • Diverse & Efficient Architectures – Offers Dense and Mixture-of-Experts (MoE) variants of different sizes for scalable deployment. • Optimized for On-Device – Smaller models are specifically designed for efficient local execution on laptops and mobile devices.
Full model card on Hugging Face →Other Gemma 4 models
Whole family →Alternatives to gemma-4-26B-A4B-it
All alternatives →Following the February release of the Qwen3.5 series, we're pleased to share the first open-weight variant of Qwen3.6. Built on direct feedback from the community, Qwen3.6 prioritizes stability and real-world utility, offering developers a more intuitive, responsive, and genuinely productive coding…
DiffusionGemma is a generative model built by Google DeepMind. Based on the 26B A4B Mixture-of-Experts (MoE) Gemma 4 architecture, DiffusionGemma generates tokens using discrete diffusion. This open-weights model is multimodal, handling text, image, and video inputs to generate text output.
Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance.
Following the February release of the Qwen3.5 series, we're pleased to share the first open-weight variant of Qwen3.6. Built on direct feedback from the community, Qwen3.6 prioritizes stability and real-world utility, offering developers a more intuitive, responsive, and genuinely productive coding…
GitHub · On-device results · Evaluation · Examples · Usage · Model weights
gemma-4-26B-A4B-it FAQ
- How much VRAM does gemma-4-26B-A4B-it need?
- About 18 GB at Q4_K_M with an 8K context. Q8_0 needs about 29 GB and full precision about 53 GB. Longer contexts need more memory for the KV cache.
- Can gemma-4-26B-A4B-it run on the Tesla P40?
- Yes. It fits at Q5_K_M (about 20 GB), at an estimated 67 tokens/s. It fits comfortably on the Radeon RX 7900 XTX, Instinct MI50 32GB, RTX 5090.
- Can I use gemma-4-26B-A4B-it commercially?
- Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.
- What is the context length of gemma-4-26B-A4B-it?
- 262,144 tokens (256K). The full window needs far more memory than an 8K context.
- How do I run gemma-4-26B-A4B-it with Ollama or llama.cpp?
- With Ollama: ollama run gemma4:26b. With llama.cpp: llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_M, which starts an OpenAI-compatible server.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments