OpenAI CEO Sam Altman said in a recent podcast teaser that Astra’s ability to operate a computer "feels fully at human level," while a separate developer finding suggests Anthropic’s Fable 5.1 has already appeared in Amazon Web Services (AWS) Bedrock API routing.
Altman ties Astra to near-AGI discussion
Only a little more than one minute of the podcast has been released so far. In the clip, the host asks: "Do you think AGI has arrived already?"
Altman answers: "Close, at least very close... AGI feels like a milestone, while superintelligence feels like an infinitely scalable monster. As for Astra, in terms of operating a computer, it feels like it has fully reached human level."
The article says several Astra demo results have surfaced recently. One test described in the report used a single prompt — "Build a gorgeous sci-fi spaceship with an explorable interior structure" — and, without multi-turn iteration, the AI generated a complete 3D world in one shot.
The source also cites a well-followed commentator who argued that AGI may not arrive in the form of a smarter chatbot, but in the form of creating a complete world from one sentence.
Astra demos highlighted in the report
The article frames Astra as a system built for "persistent agents," not as a product meant to compete with Fable on conversational skill alone.
Among the details listed, one of Astra’s latest outputs is a 3D voxel-art pagoda. At maximum reasoning intensity, the system used 56K reasoning tokens and spent 38 minutes producing a 3D structure. The report describes the result as highly polished in composition, color, and overall rendering.
The source says Astra is no longer a passive tool that responds once per prompt. It can keep working for days or even weeks, has long-term memory, and can retain user corrections.
It also says Astra can spin up multiple agents to work together on research-grade mathematical proofs. Beyond that, the report claims Astra can take over a user’s desktop at "superhuman speed," work across software tools, and collaborate with other AI agents as well as humans.
Developer says Fable 5.1 is visible in AWS Bedrock routing
On the Anthropic side, the article says a developer found that Fable 5.1 has quietly appeared in AWS Bedrock API routing.
According to the report, calls using the Fable 5.1 model identifier previously returned a 400 error, indicating an invalid model identifier. They now return a 404 error, indicating the model was not found. The article interprets that shift to mean AWS’ underlying system recognizes and accepts the Fable 5.1 identifier, even if the model itself has not yet been uploaded.
The report says this backend change matches what happened a few days before the release of Opus 5.1. It adds that more people have been included in gray testing over the past two weeks, and says these signs point to a Fable 5.1 launch window roughly 3 to 4 days away.
The article also notes that someone tested Opus 5.1 in incognito mode and found the result "quite good."
Bridgemind’s months-long testing of Fable 5 and GPT 5.6 Sol
In its model comparison section, the source cites community figure Bridgemind, who said that after months of heavy daily use of both Fable 5 and GPT 5.6 Sol, his conclusion was that GPT 5.6 was consistently beaten by Fable 5.
Based on the source article’s summary of that testing, GPT 5.6 Sol performs better on backend processing. It is faster, cheaper to call, and stronger on data security. The article also says Sol has opened a clear lead over Fable and the previous Opus generation in mathematical and logical reasoning. Once work becomes deep and complex, Sol is described as the stronger backend engine.
The same article says GPT 5.6’s weakness is "over-engineering." If instructions are even slightly loose, it may turn a simple feature into a large and complex system, forcing users to constrain it very precisely.
Fable 5, by contrast, is described in the piece as the de facto No. 1 model. Its edge, according to the article, comes from frontend UI design and one-shot execution. When given a pile of messy requirements, Fable 5 often produces something that already feels finished on the first try. The article attributes its standing among developers to that first-pass hit rate, along with strong visual output and aesthetic judgment.
Article says Astra is in limited partner testing
The source ends by saying Astra has reportedly opened limited internal testing to some partners to collect feedback. If that process goes smoothly, the article suggests users may not have to wait long to see the system more broadly.
With both Astra and Fable 5.1 generating fresh signals, the report leaves one question open: which one ships first.
References
- https://x.com/bridgemindai/status/2094406627663216733
- https://www.youtube.com/watch?v=RX26fIVOg5E
- https://x.com/pankajkumar_dev/status/2094399865971450286
The original article was published by the WeChat account "Xinzhiyuan" (ID: AI_era) and credited to Aeneas.