Many prominent AI insiders are saying they have tested Anthropic's new model and expect both releases within hours
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
Why it matters #
A cheaper GPT-6 tier would make OpenAI's newest model generation practical for coding agents and high-volume applications, but the circulated price and speed claims remain unconfirmed.
A developer who says he tested Anthropic's Opus 5.5 model now claims it and OpenAI's rumored GPT-6 Sol will arrive later today, setting up what he described as a same-day contest between the two AI companies. Neither the attached launch graphic nor the growing collection of screenshots and model identifiers establishes an official release.
Pankaj Kumar (@pankajkumar_dev) said Opus 5.5 would cost $4 per million input tokens and $20 per million output tokens, and would be closer in capability to the Fable 5.1 model. Kumar said he had tested Opus 5.5 and found it competitive with Fable 5.1 in 3D and frontend work, though not token-efficient. He also predicted that GPT-6 Sol would offer similar capability and pricing, citing the performance of OpenAI's existing GPT-6 Astra model.
The claims add a second prospective launch to the unverified screenshots and usage records already circulating on September 22nd around GPT-6 Sol. OpenAI's public model catalog, pricing page and API changelog listed no GPT-6 Sol or GPT-6 Luna as of publication.
The OpenAI claims spread through a cluster of posts from AI users watching model identifiers in Microsoft Azure and Codex sessions. KC (@karanC_12) described GPT-6 Sol as a model priced at $2.50 per million input tokens and $15 per million output tokens, with roughly six times the speed of GPT-5.6 Sol. The post also cited an early test that produced about 28,000 tokens in three minutes.
Salio (@Mr_Salio) said a ChatGPT Pro user's Codex session routed 12 requests to a model identified as gpt-6-sol, and claimed that OpenAI was also preparing GPT-6 Luna. Screenshots and model strings can indicate internal routing or limited testing, though they do not establish a release, final pricing or a stable product name. OpenAI frequently operates aliases, evaluation builds and routing layers that may never become public API products under the identifiers exposed during testing.
The rumored prices do not yet agree
Kumar's expected GPT-6 Sol price is in the same general range as his claimed $4 input and $20 output rates for Opus 5.5. That differs from the separate $2.50 input and $15 output figures circulating for GPT-6 Sol.
The lower rumored rates would place GPT-6 Sol between OpenAI's existing model tiers. OpenAI currently prices GPT-6 Astra at $10 per million input tokens and $50 per million output tokens. GPT-5.6 Sol costs $4 and $20, respectively, under promotional pricing that OpenAI says will remain available through at least November 21st.
Against Astra, the rumored $2.50/$15 GPT-6 Sol price would cut input costs by 75% and output costs by 70%. Against GPT-5.6 Sol, the reductions would be 37.5% for input and 25% for output. That math does not support the widely repeated claim that GPT-6 Sol would be about 50% cheaper than GPT-5.6 Sol at standard published rates.
The speed claim carries a similar qualification. Producing 28,000 tokens in three minutes works out to roughly 156 tokens per second, a useful rate for interactive coding and agent work. The posts provide no consistent baseline, workload, reasoning setting, context size or processing tier for the assertion that the model is six times faster. Those variables can materially change measured throughput.
OpenAI already sells faster processing separately. GPT-6 Astra's Fast mode offers up to twice the standard speed at twice the standard token rate. A model approaching the rumored throughput without a similar surcharge would give OpenAI a clearer answer for developers who find Astra capable but too expensive for routine agent loops.
OpenAI's product stack leaves room for Sol
OpenAI released GPT-6 Astra on September 3rd as its highest-capability model for coding, research, computer use and extended professional workflows. Astra has a 1.05 million-token context window, supports outputs of up to 128,000 tokens and costs substantially more than the GPT-5.6 family.
The current lineup still relies on GPT-5.6 Sol, Terra and Luna below Astra. Sol handles complex professional work, Terra balances capability and cost, and Luna targets high-volume tasks at $0.20 per million input tokens and $1.20 per million output tokens. A GPT-6 Sol would give OpenAI a mid-priced model carrying Astra-era improvements without pushing every production workload onto the $10/$50 tier.
That gap explains the intensity around a possible release. Coding agents and other long-running systems can consume large volumes of output tokens, turning small pricing differences into meaningful infrastructure costs. Astra's per-token price limits its appeal for repetitive work even when its higher capability reduces the number of attempts needed to finish a task.
OpenAI's API changelog records GPT-6 Astra as the only GPT-6 model released in September. Its latest listed update, dated September 15th, concerns API-key governance rather than a new model. The official catalog continues to recommend GPT-5.6 Terra and Luna for workloads where Astra is unnecessary.
GPT-6 Sol therefore remains an alleged internal or limited-test model, while the supplied Opus 5.5 details rest on Kumar's account of private testing. The reported positioning is commercially plausible, but the launch timing, final prices and performance claims remain unconfirmed by the companies' published materials.