Qwen3.5-35B-A3B vs Qwen3.8-27B

Qwen3.8-27B leads on the Visual Intelligence Index by 2.7 points, and Qwen3.5-35B-A3B scores higher on 1 of the 7 benchmarks where the two differ.

Model

Compared with

At a glance

Measure	Qwen3.5-35B-A3B	Qwen3.8-27B
Visual Intelligence Index	65.7%	68.5%
Rank	#7 of 9	#6 of 9
Cost per taskMean over the benchmarks, at list price.	—	—
Latency per taskEach benchmark's median, averaged.	14.3s	32.0s
Tokens per task	25,755	25,009
Input / 1M tokens	—	—
Output / 1M tokens	—	—
Context window	256K	256K
Availability	Open weights	Open weights

Each capability index is a weighted mean of the benchmarks that contribute to the capability.

Capability	Qwen3.5-35B-A3B	Qwen3.8-27B	Lead
Perception	65.3%	67.8%	−2.5
Spatial Reasoning	65.8%	68.6%	−2.9
Temporal Reasoning	65.5%	68.4%	−2.9
Action / Event Understanding	62.2%	65.2%	−2.9
Causal Reasoning	70.1%	72.3%	−2.2
Information Retrieval	63.6%	67.0%	−3.4

Benchmark	Qwen3.5-35B-A3B	Qwen3.8-27B	Lead
Video-MME v2	42.5%	50.5%	−8.1
LVBench	64.4%	64.6%	−0.2
Perception Test	66.7%	72.3%	−5.5
NExT-QA	83.4%	82.2%	+1.2
Q-Bench Video	66.9%	67.9%	−0.9
EgoSchema	77.2%	79.1%	−1.9
UCF101-AD	41.3%	45.5%	−4.2

4.5/ 5

Reachable1.0: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — supported
Portable1.0: You can run it yourself, and are not tied to one vendor.
Open weights — supported
Self-hostable — supported
Multimodal input1.0: It takes the footage directly, rather than frames you extracted.
Video ingestion — supported
Audio — supported
Programmable1.0: Output you can parse, and tools it can call on its own.
Structured output — supported
Tool calling — supported
Operable0.5: Usable interactively and in bulk, not only one call at a time.
Streaming — supported
Batch API — not supported

4.5/ 5

Reachable1.0: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — supported
Portable1.0: You can run it yourself, and are not tied to one vendor.
Open weights — supported
Self-hostable — supported
Multimodal input1.0: It takes the footage directly, rather than frames you extracted.
Video ingestion — supported
Audio — supported
Programmable1.0: Output you can parse, and tools it can call on its own.
Structured output — supported
Tool calling — supported
Operable0.5: Usable interactively and in bulk, not only one call at a time.
Streaming — supported
Batch API — not supported