Skip to content
Vision Index

GPT-6 Astra vs Kimi K3

GPT-6 Astra leads on the Visual Intelligence Index by 10.3 points, and GPT-6 Astra scores higher on 7 of the 7 benchmarks where the two differ.

OpenAIGPT-6 Astra82.2%Rank #1MoonshotKimi K371.9%Rank #5

At a glance

MeasureGPT-6 AstraKimi K3
Visual Intelligence Index82.2%71.9%
Rank#1 of 9#5 of 9
Cost per taskMean over the benchmarks, at list price.$0.864$0.114
Latency per taskEach benchmark's median, averaged.9.6s29.2s
Tokens per task86,25331,619
Input / 1M tokens$10.00$3.00
Output / 1M tokens$50.00$15.00
Context window1.05M1.05M
AvailabilityProprietaryOpen weights

How cost and latency are measured

Capabilities

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

CapabilityGPT-6 AstraKimi K3
Perception80.8%71.5%
Spatial Reasoning84.6%73.7%
Temporal Reasoning83.0%72.4%
Action / Event Understanding79.2%66.8%
Causal Reasoning83.8%75.2%
Information Retrieval80.7%70.1%

Benchmarks

BenchmarkGPT-6 AstraKimi K3
Video-MME v273.4%55.2%
LVBench81.6%66.8%
Perception Test92.5%82.4%
NExT-QA90.7%86.3%
Q-Bench Video68.3%67.5%
EgoSchema83.4%78.6%
UCF101-AD61.4%39.4%

Usability

GPT-6 Astra

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

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