Kimi K3 vs Molmo 2 8B

Kimi K3 leads on the Visual Intelligence Index by 14.3 points, and Kimi K3 scores higher on 6 of the 7 benchmarks where the two differ.

Model

Compared with

At a glance

Measure	Kimi K3	Molmo 2 8B
Visual Intelligence Index	71.9%	57.6%
Rank	#5 of 9	#9 of 9
Cost per taskMean over the benchmarks, at list price.	$0.114	—
Latency per taskEach benchmark's median, averaged.	29.2s	62.2s
Tokens per task	31,619	12,048
Input / 1M tokens	$3.00	—
Output / 1M tokens	$15.00	—
Context window	1.05M	37K
Availability	Open weights	Open weights

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

Capability	Kimi K3	Molmo 2 8B	Lead
Perception	71.5%	58.7%	+12.7
Spatial Reasoning	73.7%	61.0%	+12.7
Temporal Reasoning	72.4%	57.5%	+14.9
Action / Event Understanding	66.8%	51.1%	+15.7
Causal Reasoning	75.2%	61.7%	+13.5
Information Retrieval	70.1%	52.1%	+18.0

Benchmark	Kimi K3	Molmo 2 8B	Lead
Video-MME v2	55.2%	27.5%	+27.6
LVBench	66.8%	48.6%	+18.1
Perception Test	82.4%	67.4%	+15.0
NExT-QA	86.3%	89.2%	−3.0
Q-Bench Video	67.5%	62.9%	+4.6
EgoSchema	78.6%	58.2%	+20.4
UCF101-AD	39.4%	19.5%	+19.9

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

3.5/ 5

Reachable0.5: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — not 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