Gemma 4 26B A4B IT
- In Input modalities: TextImage Out Output modalities: Text
- Reasoning
- Reasoning can be off
- Tool use
- Structured output
- Open weights
Prices
2 offerings · prices as listedgemma-4-26b-a4b-it-maas| Conditions/ 1M tokens | Input | Cache read | Output |
|---|---|---|---|
| Standard | $0.15 | $0.015 | $0.60 |
Prices from Vertex AI generative AI pricing (opens in a new tab) (checked Nov 2, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Knowledge cutoff
- Jan 2025
- Licence
- apache-2.0
- Family
- gemma
- Model ID
Not stated on Google’s pages: max input, reasoning efforts, max thinking budget, built-in tools, aliases.
Run locally
Weights: google/gemma-4-26B-A4B-it- Parameters 25.8B
- Active 3.8B · 128 experts, 8 per token
- Native precision BF16
- Layers 30
- KV cache per token 240 KB
- Config positions 256K (config.json max_position_embeddings)
- API context 250K (maker’s API)
- Needs
- ≈ 18.5 GB (weights 17.1 + KV cache 0.5 + overhead 0.9)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 155–233 tok/s (reads the active 3.8B per token)
- Longest context
- 88K (longer doesn’t fit)
- First token
- about 0.12–0.18 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 48.1 GB | ≈ 51.0 | ○Needs system memory 13–20 tok/s | — |
| Q8_0 | 25.5 GB | ≈ 27.3 | ○Needs system memory 37–55 tok/s | — |
| FP8 | 24.0 GB | ≈ 25.7 | ○Needs system memory 43–65 tok/s | — |
| Q6_K | 19.7 GB | ≈ 21.2 | ●Runs well 137–206 tok/s | 17K |
| Q5_K_M | 17.1 GB | ≈ 18.5 | ●Runs well 155–233 tok/s | 88K |
| Q5_0 | 16.5 GB | ≈ 17.9 | ●Runs well 160–239 tok/s | 104K |
| Q4_K_M | 14.6 GB | ≈ 15.8 | ●Runs well 177–265 tok/s | 155K |
| Q4_0 | 13.5 GB | ≈ 14.7 | ●Runs well 188–282 tok/s | 184K |
| IQ4_XS | 12.8 GB | ≈ 13.9 | ●Runs well 197–295 tok/s | 204K |
| Q3_K_Mquality drops | 11.7 GB | ≈ 12.8 | ●Runs well 210–315 tok/s | 232K |
| IQ3_XXSquality drops | 9.2 GB | ≈ 10.2 | ●Runs well 252–378 tok/s | 250K |
| Q2_Kquality drops | 9.0 GB | ≈ 10.0 | ●Runs well 256–384 tok/s | 250K |
| IQ2_XXSquality drops | 6.2 GB | ≈ 7.2 | ●Runs well 331–496 tok/s | 250K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 4090; unified memory Apple M5 Pro (16-core GPU) · 24 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- Epoch Capabilities Index 141.7
#92 of 118 listed models · CI 139.2–144.1
- Epoch AI · GPQA Diamond 73.2%
#92 of 168 listed models · minimal effort · ±3.15
- Epoch AI · OTIS Mock AIME 2024–2025 82.2%
#78 of 159 listed models · minimal effort · ±5.76
- EQ-Bench Creative Writing v3 1,305
#73 of 93 listed models
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Nov 4 – 5, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 28, 2026
- changes we made
Changes
All changes →- Now priced on Cloudflare Workers AI Input $0.10 · Output $0.30
- Now priced on Google Cloud Vertex AI Input $0.15 · Output $0.60 · Cache read $0.015
Sources
- Gemma 4 model card (opens in a new tab) ai.google.dev · checked Nov 1, 2026
- Gemini API changelog (opens in a new tab) ai.google.dev · checked Nov 1, 2026
- Vertex AI generative AI pricing (opens in a new tab) cloud.google.com · checked Nov 2, 2026
- Gemma 4 26B A4B IT | Gemini Enterprise Agent Platform | Google Cloud Documentation (opens in a new tab) docs.cloud.google.com · checked Nov 1, 2026
- Thinking for open models | Gemini Enterprise Agent Platform | Google Cloud Documentation (opens in a new tab) docs.cloud.google.com · checked Nov 4, 2026
- Cloudflare Workers AI pricing (opens in a new tab) developers.cloudflare.com · checked Oct 31, 2026
- google/gemma-4-26B-A4B-it model card (opens in a new tab) huggingface.co · checked Nov 3, 2026
- google/gemma-4-26B-A4B-it config.json (opens in a new tab) huggingface.co · checked Nov 3, 2026
- google/gemma-4-26B-A4B-it files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- google/gemma-4-26B-A4B-it README.md (opens in a new tab) huggingface.co · checked Nov 1, 2026
- google/gemma-4-26B-A4B-it model card (opens in a new tab) huggingface.co · checked Nov 3, 2026
Google’s official policies: data use, privacy, terms →
Google · United States · ai.google.dev
Related models
Snapshots, the same family and other models from this provider