Qwen3.5-35B-A3B

Qwen3.5-35B-A3B is a multimodal Mixture-of-Experts model with 35 billion total parameters and 3 billion activated parameters. It combines strong reasoning, coding, agentic, and visual understanding performance with production-friendly efficiency and a native 262K context window.

CountBench

97.8%

i
VLMsAreBlind

97.0%

i
MMLU-Redux

93.3%

i
V*

92.7%

i
AI2D

92.6%

i
IFEval

91.9%

i
MMBench-V1.1

91.5%

i
OCRBench

91.0%

i
C-Eval

90.2%

i
OmniDocBench 1.5

89.3%

i
HMMT25

89.2%

i
RefCOCO-avg

89.2%

i
HMMT 2025

89.0%

i
Global PIQA

86.6%

i
MAXIFE

86.6%

i
VideoMME w sub.

86.6%

i
MathVista-Mini

86.2%

i
MLVU

85.6%

i
MMLU-Pro

85.3%

i
MMMLU

85.2%

i
DynaMath

85.0%

i
GPQA

84.2%

i
RealWorldQA

84.1%

i
MathVision

83.9%

i
EmbSpatialBench

83.1%

i
VideoMME w/o sub.

82.5%

i
CodeForces

82.2%

i
MMStar

81.9%

i
MMMU

81.4%

i
t2-bench

81.2%

i
MMLU-ProX

81.0%

i
CC-OCR

80.7%

i
VideoMMMU

80.4%

i
Include

79.7%

i
LingoQA

79.2%

i
SlakeVQA

78.7%

i
CharXiv-R

77.5%

i
WMT24++

76.3%

i
MMMU-Pro

75.1%

i
MVBench

74.8%

i
LiveCodeBench v6

74.6%

i
MMVU

72.3%

i
LVBench

71.4%

i
AndroidWorld_SR

71.1%

i
IFBench

70.2%

i
BrowseComp-zh

69.5%

i
SWE-Bench Verified

69.2%

i
ScreenSpot Pro

68.6%

i
Hallusion Bench

67.9%

i
BFCL-V4

67.3%

i
ERQA

64.8%

i
PolyMATH

64.4%

i
RefSpatialBench

63.5%

i
SuperGPQA

63.4%

i
PMC-VQA

62.0%

i
MedXpertQA

61.4%

i
BrowseComp

61.0%

i
Multi-Challenge

60.0%

i
MMLongBench-Doc

59.5%

i
LongBench v2

59.0%

i
AA-LCR

58.5%

i
SimpleVQA

58.3%

i
FullStackBench en

58.1%

i
NOVA-63

57.1%

i
WideSearch

57.1%

i
TIR-Bench

55.5%

i
FullStackBench zh

55.0%

i
OSWorld-Verified

54.5%

i
Humanity's Last Exam

47.4%

i
ODinW

42.6%

i
Seal-0

41.4%

i
Terminal-Bench 2.0

40.5%

i
BabyVision

38.4%

i
OJBench

36.0%

i
ZEROBench-Sub

34.1%

i
SUNRGBD

33.4%

i
VITA-Bench

31.9%

i
DeepPlanning

22.8%

i
Nuscene

14.6%

i
Hypersim

13.1%

i
ZEROBench

8.0%

i

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

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
AkashML	available	$0.14/Mtok	$1.00/Mtok	262K tokens context 262K tokens max output	100.0% 5m 100.0%	830 ms p50 TTFT 97 tok/s p50	fp8
DeepInfra	available	$0.14/Mtok cache $0.05/Mtok	$1.00/Mtok	262K tokens context 82K tokens max output	100.0% 5m 100.0%	486 ms p50 TTFT 100 tok/s p50	fp8
Parasail	available	$0.15/Mtok cache $0.05/Mtok	$1.00/Mtok	262K tokens context 262K tokens max output	100.0% 5m 100.0%	1,889 ms p50 TTFT 32 tok/s p50	fp8
Alibaba	available	$0.16/Mtok	$1.30/Mtok	262K tokens context 66K tokens max output	100.0% 5m 100.0%	618 ms p50 TTFT 81 tok/s p50
AtlasCloud	available	$0.22/Mtok cache $0.22/Mtok	$1.80/Mtok	262K tokens context 66K tokens max output	100.0% 5m 100.0%	2,646 ms p50 TTFT 32 tok/s p50	fp8
NextBit	available	$0.23/Mtok	$1.60/Mtok	262K tokens context 262K tokens max output	100.0% 5m 100.0%	1,948 ms p50 TTFT 37 tok/s p50	fp8
SiliconFlow	available	$0.24/Mtok	$1.80/Mtok	262K tokens context 262K tokens max output	96% 5m 93%	4,802 ms p50 TTFT 16 tok/s p50	fp8
WandB	available	$0.25/Mtok cache $0.25/Mtok	$1.25/Mtok	262K tokens context 262K tokens max output	99% 5m 97%	924 ms p50 TTFT 83 tok/s p50	fp8
Venice	available	$0.31/Mtok cache $0.16/Mtok	$1.25/Mtok	256K tokens context 16K tokens max output	100.0%	3,122 ms p50 TTFT 28 tok/s p50