Gemini 3.8 Flash vs Kimi K3

Gemini 3.8 Flash leads on the Visual Intelligence Index by 8.7 points, and Gemini 3.8 Flash scores higher on 7 of the 7 benchmarks where the two differ.

Model

Compared with

GoogleGemini 3.8 Flash80.6%Rank #2 MoonshotKimi K371.9%Rank #5

At a glance

Measure	Gemini 3.8 Flash	Kimi K3
Visual Intelligence Index	80.6%	71.9%
Rank	#2 of 9	#5 of 9
Cost per taskMean over the benchmarks, at list price.	$0.0474	$0.114
Latency per taskEach benchmark's median, averaged.	25.5s	29.2s
Tokens per task	60,794	31,619
Input / 1M tokens	$0.75	$3.00
Output / 1M tokens	$3.75	$15.00
Context window	1M	1.05M
Availability	Proprietary	Open weights

How cost and latency are measured

Capabilities

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

Capability	Gemini 3.8 Flash	Kimi K3	Lead
Perception	79.8%	71.5%	+8.3
Spatial Reasoning	82.5%	73.7%	+8.8
Temporal Reasoning	81.3%	72.4%	+8.9
Action / Event Understanding	78.2%	66.8%	+11.4
Causal Reasoning	81.6%	75.2%	+6.4
Information Retrieval	79.4%	70.1%	+9.2

Benchmarks

Benchmark	Gemini 3.8 Flash	Kimi K3	Lead
Video-MME v2	71.3%	55.2%	+16.1
LVBench	86.4%	66.8%	+19.6
Perception Test	85.5%	82.4%	+3.1
NExT-QA	88.0%	86.3%	+1.7
Q-Bench Video	69.3%	67.5%	+1.8
EgoSchema	80.8%	78.6%	+2.2
UCF101-AD	63.5%	39.4%	+24.1

Usability

Gemini 3.8 Flash

4.0/ 5

Reachable1.0: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — supported
Portable0.0: You can run it yourself, and are not tied to one vendor.
Open weights — not supported
Self-hostable — not 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
Operable1.0: Usable interactively and in bulk, not only one call at a time.
Streaming — supported
Batch API — supported

Kimi K3

4.0/ 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 input0.5: It takes the footage directly, rather than frames you extracted.
Video ingestion — supported
Audio — not 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