Astra 6.2
An AI model equipped with an advanced reasoning mode that performs strongly in visual content generation but exhibits issues with content accuracy and hallucination.
Timeline (last 90 days)
Bilibili uploader "Lei Ge" compared Gemini 4 Neon and Astra 6.2 on making a PPT from the same prompt.
Astra 6.2 used "advanced reasoning" mode for the PPT-making task, taking more than four minutes to generate.
Astra 6.2 generated a 16-page PPT in the PPT task.
The uploader rated the visual effect of the PPT generated by Neon 4 as no worse than Astra 6.2's, and more concise and intuitive.
On the page "The difference between AI digital employees and traditional outbound calling," the content Astra 6.2 generated did not match the provided outline at all, and was called "nonsense" by the uploader.
Astra completed the teardown of the iPhone Duo foldable screen in 65 minutes: it first extracted 73 meshes from the official model, found the inner-screen assembly positioning diagram and precise dimensions in the repair manual, and generated 7 GLB models of the screen
Astra 6.2 cuts prices right at launch: input and output each drop 20%, cache reads drop 60%
The uploader suggests that for long-term projects, first use Astra to build the overall framework, then hand local refinement and revisions to Sol to implement
In the Chinese writing test, Astra failed badly: the whole story had no main thread, the three characters were unrelated to each other, and the plot went its separate ways
After the price cut, Astra 6.2's unit price is still twice that of Claude Opus 5.5
3 more timeline events
All Facts (13)
On the page "The difference between AI digital employees and traditional outbound calling," the content Astra 6.2 generated did not match the provided outline at all, and was called "nonsense" by the uploader.
50%UnverifiedAstra 6.2 used "advanced reasoning" mode for the PPT-making task, taking more than four minutes to generate.
50%UnverifiedAstra 6.2 generated a 16-page PPT in the PPT task.
50%UnverifiedThe uploader rated the visual effect of the PPT generated by Neon 4 as no worse than Astra 6.2's, and more concise and intuitive.
50%UnverifiedIn the test of generating a 3D scene of Beijing's first snowfall, Astra 6.2 took one hour and eighteen minutes, the longest among all the contestants
50%UnverifiedAstra completed the teardown of the iPhone Duo foldable screen in 65 minutes: it first extracted 73 meshes from the official model, found the inner-screen assembly positioning diagram and precise dimensions in the repair manual, and generated 7 GLB models of the screen
50%UnverifiedAstra 6.2 cuts prices right at launch: input and output each drop 20%, cache reads drop 60%
50%UnverifiedThe uploader suggests that for long-term projects, first use Astra to build the overall framework, then hand local refinement and revisions to Sol to implement
50%UnverifiedIn the Chinese writing test, Astra failed badly: the whole story had no main thread, the three characters were unrelated to each other, and the plot went its separate ways
50%UnverifiedAfter the price cut, Astra 6.2's unit price is still twice that of Claude Opus 5.5
50%UnverifiedAccording to the third-party organization Artificial Analysis's comprehensive intelligence leaderboard, Astra 6.2 ranks second, behind Fable 5.2
50%UnverifiedOn the Coding Agent programming leaderboard, Astra 6.2 is exactly on par with the previous generation, while Sol 6.2 dropped by three points compared with the previous generation
50%UnverifiedBilibili uploader "Lei Ge" compared Gemini 4 Neon and Astra 6.2 on making a PPT from the same prompt.
50%