Gemma 4 26B-A4B

Gemma 4 26B-A4B is Google DeepMind's Mixture-of-Experts multimodal model with 26 billion total parameters and 3.8 billion activated parameters per token, with a 256K context window. Ranks #6 among open models on Arena AI, outcompeting models 20x its size.

AIME 2026

88.3%

i
MMMLU

86.3%

i
t2-bench

85.5%

i
MMLU-Pro

82.6%

i
MathVision

82.4%

i
GPQA

82.3%

i
LiveCodeBench v6

77.1%

i
MMMU-Pro

73.8%

i
BIG-Bench Extra Hard

64.8%

i
MedXpertQA

58.1%

i
MRCR v2

44.1%

i
Humanity's Last Exam

17.2%

i

Pricing, uptime, and speed via OpenRouter — updated Jul 17, 2026, 04:19 AM.

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
DeepInfra	available	$0.07/Mtok	$0.34/Mtok	262K tokens context 16K tokens max output	99.9% 5m 100.0%	498 ms p50 TTFT 30 tok/s p50	fp8
Cloudflare	available	$0.10/Mtok	$0.30/Mtok	256K tokens context 256K tokens max output	99% 5m 99%	341 ms p50 TTFT 38 tok/s p50
NextBit	available	$0.12/Mtok cache $0.05/Mtok	$0.35/Mtok	262K tokens context 262K tokens max output	99.8% 5m 99.9%	3,679 ms p50 TTFT 12 tok/s p50	bf16
Novita	available	$0.13/Mtok	$0.40/Mtok	262K tokens context 131K tokens max output	99.1% 5m 98%	1,444 ms p50 TTFT 26 tok/s p50	bf16
Parasail	available	$0.13/Mtok cache $0.05/Mtok	$0.40/Mtok	262K tokens context 262K tokens max output	99% 5m 98%	773 ms p50 TTFT 28 tok/s p50	bf16
Venice	available	$0.13/Mtok cache $0.05/Mtok	$0.40/Mtok	256K tokens context 8K tokens max output	99% 5m 99.2%	1,065 ms p50 TTFT 24 tok/s p50	bf16
Wafer	available	$0.13/Mtok cache $0.05/Mtok	$0.40/Mtok	262K tokens context 16K tokens max output	100.0% 5m 100.0%	658 ms p50 TTFT 78 tok/s p50	fp8
DekaLLM	-2	$0.06/Mtok	$0.33/Mtok	262K tokens context 33K tokens max output	90% 5m 91%	1,434 ms p50 TTFT 43 tok/s p50	bf16
SiliconFlow	-5	$0.12/Mtok	$0.40/Mtok	262K tokens context 262K tokens max output	25% 5m 61%	5,363 ms p50 TTFT 15 tok/s p50	fp8
Google	-5	$0.15/Mtok	$0.60/Mtok	262K tokens context 262K tokens max output	41% 5m 33%	375 ms p50 TTFT 32 tok/s p50