GPT-6 Astra vs Gemini 4 Argon

Gemini 4 Argon leads on the Visual Intelligence Index by 2.0 points, and GPT-6 Astra scores higher on 3 of the 7 benchmarks where the two differ.

Model

Compared with

At a glance

Measure	GPT-6 Astra	Gemini 4 Argon
Visual Intelligence Index	82.2%	84.1%
Rank	#2 of 10	#1 of 10
Cost per taskMean over the benchmarks, at list price.	$0.864	$0.110
Latency per taskEach benchmark's median, averaged.	9.6s	23.9s
Tokens per task	86,253	54,614
Input / 1M tokens	$10.00	$2.00
Output / 1M tokens	$50.00	$10.00
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	GPT-6 Astra	Gemini 4 Argon	Lead
Perception	80.8%	81.7%	−0.9
Spatial Reasoning	84.6%	85.4%	−0.9
Temporal Reasoning	83.0%	84.5%	−1.5
Action / Event Understanding	79.2%	84.7%	−5.5
Causal Reasoning	83.8%	85.4%	−1.7
Information Retrieval	80.7%	82.7%	−2.0

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

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

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