Three Unlisted Model Codenames — 'GPT Terra', 'GPT Sol', 'GPT Astra' — Surface on OpenRouter, Not Confirmed by OpenAI
Three model identifiers — gpt-terra-latest, gpt-sol-latest, and gpt-astra-latest — appeared as OpenRouter listing pages under an '~openai' namespace, but no official OpenAI announcement, specs, or pricing accompany them. The listings' authenticity and meaning remain unverified.
What surfaced
Three URLs on OpenRouter — a third-party model marketplace and API router — reference model identifiers under a ~openai namespace: gpt-terra-latest, gpt-sol-latest, and gpt-astra-latest. Each has its own listing page on OpenRouter's model catalog. No press release, blog post, or API changelog from OpenAI accompanies any of these names as of this writing.
The tilde-prefixed ~openai path structure on OpenRouter is unusual and does not match the standard openai/ namespace the platform uses for verified OpenAI models such as GPT-4o or GPT-4.1. This distinction matters: OpenRouter allows community-contributed or proxy listings that are not always vetted, confirmed, or operated directly by the named company.
What's confirmed — and what isn't
No context window size, pricing per token, benchmark scores, or training cutoff date has been published for any of the three names. No parameter count has been disclosed. There is no statement from OpenAI acknowledging Terra, Sol, or Astra as product names, internal codenames, or upcoming releases.
What can be said with certainty: the listing pages exist on OpenRouter's site under those URLs. Beyond that, everything else — whether these represent real forthcoming OpenAI models, placeholder entries, test scaffolding, third-party misattribution, or an unrelated party's naming scheme — is unverified.
Celestial-body naming (Terra, Sol, Astra) has not previously been used by OpenAI for any released model; the company's public naming conventions have followed GPT-n version numbers (GPT-3.5, GPT-4, GPT-4o, GPT-4.1, GPT-5) or descriptive suffixes (o1, o3, mini, Turbo). That departure is itself a reason for caution rather than confirmation.
Why this keeps happening
OpenRouter and similar aggregator platforms frequently surface model IDs before or independent of an official vendor announcement, sometimes because a partner integration is being tested, sometimes because a listing is speculative or auto-generated from leaked API references, and occasionally because of outright naming errors in a marketplace's backend. Historically, some of these early listings have preceded genuine releases by days or weeks; others have never resolved into anything shipped.
What this means
Treat "GPT Terra," "GPT Sol," and "GPT Astra" as unconfirmed until OpenAI publishes model documentation, pricing, and benchmark data through its own channels (the API reference, help center, or official announcement). Builders should not architect production systems around these identifiers yet — there is no guarantee of API stability, no published rate limits, and no confirmation the endpoints will remain live under these names. The responsible move is to monitor OpenAI's official channels directly rather than treat third-party marketplace listings as a launch signal. If and when OpenAI confirms any of these models, expect a follow-up with concrete specs.
Related Articles
Three Unverified 'GPT-6' Model Listings Appear on OpenRouter: Sol Pro, Luna, and Luna Pro
Three model pages bearing the names GPT-6 Sol Pro, GPT-6 Luna, and GPT-6 Luna Pro have surfaced on OpenRouter's site, but OpenAI has issued no official announcement confirming these as real releases. No pricing, benchmark scores, or context window figures have been disclosed.
Xiaomi Lists Three New MiMo-V2.6 Models on OpenRouter: Pro, Flash, and Pro-UltraSpeed
Xiaomi has added three new entries to its MiMo model family on OpenRouter: MiMo-V2.6-Pro, MiMo-V2.6-Flash, and MiMo-V2.6-Pro-UltraSpeed. Full specifications, pricing, and benchmark results have not yet been disclosed.
Google and Meta unveil frontier models they claim match OpenAI and Anthropic's best
Google and Meta each unveiled a new frontier model within a span of days, and both companies claim the models can compete with the best from OpenAI and Anthropic. Independent benchmarks, pricing and technical specifications were not available in the source reporting.
Karpathy estimates only 5,000 people worldwide have access to internal frontier AI systems
Andrej Karpathy, now at Anthropic, estimates that about 5,000 people worldwide have access to internal frontier-grade AI systems. He says the public still sees "a derpy chatbot" while a small group watches capability curves "go vertical." The figures are his own estimates, not measured data.
Comments
Loading...