Gemini 4 Argon vs GPT-6 Astra

Gemini 4 Argon leads on the Visual Intelligence Index by 2.0 points, and Gemini 4 Argon scores higher on 4 of the 7 benchmarks where the two differ.

Model

Compared with

At a glance

Measure	Gemini 4 Argon	GPT-6 Astra
Visual Intelligence Index	84.1%	82.2%
Rank	#1 of 10	#2 of 10
Cost per taskMean over the benchmarks, at list price.	$0.110	$0.864
Latency per taskEach benchmark's median, averaged.	23.9s	9.6s
Tokens per task	54,614	86,253
Input / 1M tokens	$2.00	$10.00
Output / 1M tokens	$10.00	$50.00
Context window	1.05M	1.05M
Availability	Proprietary	Proprietary

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

Capability	Gemini 4 Argon	GPT-6 Astra	Lead
Perception	81.7%	80.8%	+0.9
Spatial Reasoning	85.4%	84.6%	+0.9
Temporal Reasoning	84.5%	83.0%	+1.5
Action / Event Understanding	84.7%	79.2%	+5.5
Causal Reasoning	85.4%	83.8%	+1.7
Information Retrieval	82.7%	80.7%	+2.0

Benchmark	Gemini 4 Argon	GPT-6 Astra	Lead
Video-MME v2	77.0%	73.4%	+3.6
LVBench	86.3%	81.6%	+4.7
Perception Test	89.7%	92.5%	−2.9
NExT-QA	90.0%	90.7%	−0.7
Q-Bench Video	66.9%	68.3%	−1.4
EgoSchema	86.4%	83.4%	+3.0
UCF101-AD	81.8%	61.4%	+20.4

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

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