Gemma 3 27B

Gemma 3 27B is a 27-billion-parameter vision-language model from Google, handling text and image input and generating text output. It features a 128K context window, multilingual support, and open weights.

GSM8k

95.9%

i
IFEval

90.4%

i
MATH

89.0%

i
HumanEval

87.8%

i
BIG-Bench Hard

87.6%

i
DocVQA

86.6%

i
AI2D

84.5%

i
Natural2Code

84.5%

i
ChartQA

78.0%

i
Global-MMLU-Lite

75.1%

i
FACTS Grounding

74.9%

i
MBPP

74.4%

i
VQAv2 (val)

71.0%

i
InfoVQA

70.6%

i
MathVista-Mini

67.6%

i
MMLU-Pro

67.5%

i
TextVQA

65.1%

i
MMMU (val)

64.9%

i
HiddenMath

60.3%

i
Bird-SQL (dev)

54.4%

i
WMT24++

53.4%

i
GPQA

42.4%

i
LiveCodeBench

29.7%

i
BIG-Bench Extra Hard

19.3%

i
ECLeKTic

16.7%

i
SimpleQA

10.0%

i

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

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
DeepInfra	available	$0.08/Mtok	$0.16/Mtok	131K tokens context 16K tokens max output	100.0% 5m 100.0%	706 ms p50 TTFT 19 tok/s p50	fp8
Parasail	available	$0.08/Mtok cache $0.04/Mtok	$0.45/Mtok	131K tokens context 131K tokens max output	99.9% 5m 100.0%	607 ms p50 TTFT 37 tok/s p50	fp8
Novita	available	$0.12/Mtok	$0.20/Mtok	98K tokens context 16K tokens max output	98% 5m 92%	598 ms p50 TTFT 26 tok/s p50	bf16
Nebius	-2	$0.10/Mtok	$0.30/Mtok	110K tokens context	94% 5m 87%	801 ms p50 TTFT 19 tok/s p50	fp8
Phala	-5	$0.15/Mtok cache $0.07/Mtok	$0.46/Mtok	262K tokens context 262K tokens max output	63% 5m 53%	1,923 ms p50 TTFT 12 tok/s p50