Google · Benchmarks

Gemma 4 26B A4B benchmarks

Gemma 4 26B A4B has 9 published scores in this dataset, from a Google release dated 2026-04-02. Its composite index is 72.9, which places it 19th of 22 ranked models and 14th among 15 open-weight models. Its strongest area is knowledge (index 89.3) and its weakest is agentic (index 56.6). Its best placing is on LMArena Text Elo, where its 1438 ranks 17th of 21 models with a published score.

Deploy hereDeploy Gemma 4 26B A4BOfficial page

You can self-deploy Gemma 4 26B A4B on a dedicated GPU with OpenLLM Buddy and call it through an OpenAI-compatible API.

Category profile

Index 0–100: each score relative to the best published score on that test.

ReasoningCodingAgenticKnowledge
Reasoning
63.2#16 of 16 · 3 tests
Coding
82.51 test · not ranked
Agentic
56.6#7 of 8 · 2 tests
Knowledge
89.3#4 of 4 · 2 tests

Model facts

Lab
Google
Type
OpenLLM Buddy catalog
Released
2026-04-02
Parameters
25.2B total, 3.8B active (MoE)
Context
256K
License
Apache-2.0

Gemma 4 26B A4B hardware requirements and cost

At 4-bit it needs about 17 GB, which fits a single RTX 4090, so it can run on a home workstation with llama.cpp, Ollama, or vLLM. Full BF16 precision needs about 58 GB (1× H100 (80 GB)).

Estimated GPU memory for Gemma 4 26B A4B
PrecisionWeightsWith headroomSmallest setup
BF16 (full precision)50 GB58 GB1× H100 (80 GB)
FP8 / 8-bit25 GB29 GBRTX 5090 (32 GB)
4-bit (GGUF Q4_K_M)15 GB17 GBRTX 4090 (24 GB)

Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.

On OpenLLM Buddy: Gemma 4 26B A4B, with an OpenAI-compatible API and no per-token fees. Prices start at $13 USD for 11 Hours.

  • 11 Hours: $13 USD
  • 24 Hours: $28 USD
  • 1 Week: $192 USD
  • 1 Month: $767 USD
Deploy Gemma 4 26B A4B

Lighter or stronger open models

Open models ranked just above and below on the composite index, with their 4-bit memory needs.

Gemma 4 26B A4B speed and pricing

Measured figures link to their source; estimates are labelled.

API output speed
113 tok/sArtificial Analysis median · Artificial Analysis
Time to first token
0.53 sOpenRouter model page provider stats
API price per 1M tokens
$0.150 in · $0.600 outlab's own endpoint on OpenRouter · OpenRouter endpoints API
On your own RTX 4090 (4-bit)estimate
133–160 tok/s

Estimate for one user on one GPU: active weights read per token ÷ memory bandwidth at 50–70% efficiency, plus 3 ms overhead per token. Real speed varies with engine, context length and quantization; serving several users at once raises total throughput.

Compare speed and price across all models

Score by test

Rank is among every model with a published score on that test.

GPQA Diamond82.3%Graduate-level, Google-proof science questions in biology, physics, and chemistry.#18 of 19 · 13.8 behind GPT-6 AstraHumanity's Last Exam8.7%no tools#23 of 23 · 55.7 behind Claude Opus 5.5Humanity's Last Exam (with tools)17.2%with search#17 of 17 · 52.8 behind Kimi K3.1 ProAIME88.3%AIME 2026, no tools#5 of 5 · 9.4 behind Reflection Beam 1.0LiveCodeBench77.1%LiveCodeBench v6#7 of 7 · 16.4 behind DeepSeek V4 Proτ²-bench68.2%avg. over 3 domains#2 of 2 · 8.7 behind Gemma 4 31BMMLU-Pro82.6%Broad academic knowledge with ten answer options instead of four.#5 of 5 · 4.9 behind DeepSeek V4 ProMMMU-Pro73.8%MMMU-Pro#9 of 9 · 13.9 behind Claude Opus 5.5LMArena Text Elo1438LMArena text overall, 'Gemma 4 26B A4B', ±7, as of 2026-10-27#17 of 21 · 82 behind Gemini 4 Argon

Nearby in the composite ranking

Models with the closest composite index.

  1. 1Gemma 4 31B79.2
  2. 2Step 5 Preview78.7
  3. 3Nemotron 3 Ultra74.4
  4. 4Gemma 4 26B A4B72.9
  5. 5Claude Haiku 5.572.4
  6. 6Gemini 3.8 Flash70.8
  7. 7gpt-oss-120b57.3

Head-to-head comparisons

Side-by-side pages built from the tests both models report.

Sources

ModelTestScoreSettingSource
Gemma 4 26B A4BGPQA Diamond82.3—Gemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BHumanity's Last Exam8.7no toolsGemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BHumanity's Last Exam (with tools)17.2with searchGemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BAIME88.3AIME 2026, no toolsGemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BLiveCodeBench77.1LiveCodeBench v6Gemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4Bτ²-bench68.2avg. over 3 domainsGemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BMMLU-Pro82.6—Gemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BMMMU-Pro73.8MMMU-ProGemma 4 31B IT model card (Hugging Face), 26B A4B column
Gemma 4 26B A4BLMArena Text Elo1438LMArena text overall, 'Gemma 4 26B A4B', ±7, as of 2026-10-27Arena (LMArena) text leaderboard

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works