GPT-6 Astra vs Claude Fable 5.1

GPT-6 Astra leads on the Visual Intelligence Index by 6.0 points, and GPT-6 Astra scores higher on 6 of the 7 benchmarks where the two differ.

Model

Compared with

At a glance

Measure	GPT-6 Astra	Claude Fable 5.1
Visual Intelligence Index	82.2%	76.2%
Rank	#1 of 9	#3 of 9
Cost per taskMean over the benchmarks, at list price.	$0.864	$0.517
Latency per taskEach benchmark's median, averaged.	9.6s	9.0s
Tokens per task	86,253	51,151
Input / 1M tokens	$10.00	$10.00
Output / 1M tokens	$50.00	$50.00
Context window	1.05M	1M
Availability	Proprietary	Proprietary

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

Capability	GPT-6 Astra	Claude Fable 5.1	Lead
Perception	80.8%	75.3%	+5.5
Spatial Reasoning	84.6%	77.0%	+7.5
Temporal Reasoning	83.0%	76.4%	+6.6
Action / Event Understanding	79.2%	74.0%	+5.3
Causal Reasoning	83.8%	78.7%	+5.0
Information Retrieval	80.7%	74.7%	+6.0

Benchmark	GPT-6 Astra	Claude Fable 5.1	Lead
Video-MME v2	73.4%	66.2%	+7.2
LVBench	81.6%	75.0%	+6.6
Perception Test	92.5%	77.3%	+15.2
NExT-QA	90.7%	86.8%	+3.9
Q-Bench Video	68.3%	71.0%	−2.7
EgoSchema	83.4%	80.4%	+3.0
UCF101-AD	61.4%	60.6%	+0.8

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

3.0/ 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 input0.0: It takes the footage directly, rather than frames you extracted.
Video ingestion — not supported
Audio — not 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