Gemma 4 26B A4B
Gemma 4 26B A4BRankingin non-thinking mode?
Alternatives?
CheaperMiMo v2.5thinkingOpen+67%-33%$0.019Memory (4-bit): 15 GB → 186 GBOpenGemma 4 31Bthinking+45%-39%$0.10 / $0.34Memory (4-bit): 15 GB → 19 GBCheaperMiMo v2 Omnithinking+39%-33%$0.14 / $0.28
CheaperQwen 3.5 9BthinkingOpensame-65%$0.21
CheaperGPT OSS 120BhighOpen-6%-77%$0.092Context: 256K → 131K+2MiMo v2.5thinkingOpen+67%-33%$0.019Memory (4-bit): 15 GB → 186 GBGemma 4 31BthinkingOpen+45%-39%$0.10 / $0.34Memory (4-bit): 15 GB → 19 GBMiMo v2 Omnithinking+39%-33%$0.14 / $0.28Qwen 3.5 9BthinkingOpensame-65%$0.21GPT OSS 120BhighOpen-6%-77%$0.092Context: 256K → 131K+2MiMo v2.5thinkingOpen+67%-33%$0.019Memory (4-bit): 15 GB → 186 GBGPT-5 Minihigh+33%+162%$0.053MiMo v2.5 ProthinkingOpen+73%+107%$0.054Memory (4-bit): 15 GB → 614 GB+1GPT-5.6 Terralow+76%+1,614%$0.14Prompts with >272K input tokensGPT-5.4 Nanoxhigh+70%+76%$0.18Gemma 4 31BthinkingOpen+45%-39%$0.10 / $0.34Memory (4-bit): 15 GB → 19 GBQwen3.6 27BthinkingOpen+39%+414%$0.62Memory (4-bit): 15 GB → 17 GBQwen3.6 35B A3BthinkingOpen+39%+112%$0.48Memory (4-bit): 15 GB → 22 GBQwen 3.5 27BthinkingOpen+30%+214%$0.30 / $2.40Memory (4-bit): 15 GB → 17 GB
North Mini CodeOpen+24%——Memory (4-bit): 15 GB → 18 GB+1Gemini 3.1 Pro Previewthinking×2.2—$1.30Cannot use both thinking controlsGemma 4 31BthinkingOpen+45%—$0.10 / $0.34Memory (4-bit): 15 GB → 19 GBGemini 3 Flash Previewnon-thinking+27%—$0.50 / $3Gemini 2.5 Prothinking+6%—$0.33Gemini 3.1 Flash Litehighsame—$0.06Gemma 4 12BthinkingOpen-27%—$0.10 / $0.30Gemini 2.5 Flash Preview (Sep '25)non-thinking-42%——Gemini 2.5 Flashnon-thinking-52%—$0.30 / $2.50All Google models →Run it yourself
estimateBenchmarks
all categoriesFAQ
How much does Gemma 4 26B A4B cost?›
$0.15 per 1M input tokens and $0.60 per 1M output tokens; cheapest via Darkbloom at $0.042 / $0.22.
Is Gemma 4 26B A4B free to use?›
There is no free API; the cheapest paid price is $0.042 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Gemma 4 26B A4B, and what is it best at?›
#166 of 334 in Chat & Knowledge, #78 of 97 in Coding, #99 of 133 in Agents & Tool Use, #70 of 195 in Writing & Language, #77 of 124 in Vision & Documents.
Is Gemma 4 26B A4B better than DeepSeek V3.1?›
On Terminal-Bench Hard (version unstated), Gemma 4 26B A4B scores 25 vs 25 — about the same.
How can I use Gemma 4 26B A4B?›
Through the API of Cloudflare, CoreWeave, Darkbloom, DeepInfra, DekaLLM and 8 more.
What is the context window of Gemma 4 26B A4B?›
256K tokens, with up to 236K output tokens.
When was Gemma 4 26B A4B released? Is there a newer version?›
Released 2026-04-02. It is the current version in its line.
Is Gemma 4 26B A4B open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Gemma 4 26B A4B locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Gemma 4 26B A4B read images?›
It accepts text and images, video as input.
What is Gemma 4 26B A4B?›
Gemma 4 26B A4B is an open-weights language model from Google, released Apr 2026.
What are the best alternatives to Gemma 4 26B A4B?›
MiMo v2.5 (cheaper), Gemma 4 31B (open weights).