OpenAI and Anthropic Dropped New Models on the Same Day — and Both Moved Toward Cheaper, Broader Access

The same-day pattern

Two frontier labs dropped new models on the same calendar day, and the releases shared a theme that is hard to ignore: the next move in frontier AI is not a bigger number on a benchmark — it is cheaper, more accessible, and pointed at a wider slice of users and developers.

OpenAI launched GPT-6 Sol and GPT-6 Luna on September 22, 2026, and Anthropic launched Claude Opus 5.5 the same afternoon. Sol is OpenAI's new mid-tier, sitting below the flagship GPT-6 Astra and above the low-tier Luna, priced at $2 per million input tokens and $10 per million output tokens with cached input reads at $0.20 per million. It is built for multistep work — agentic coding, long-running tasks — the kind of workload that wants GPT-6 reasoning without paying Astra prices for everything, and it runs on a 1,050,000-token context window with a maximum output of 128,000 tokens. On Artificial Analysis's numbers, Sol scores around 82.2 on the Decision benchmark and ranks fourth out of 196 models tracked, with its strongest eligible category in coding. Luna is the low end of the GPT-6 stack at $0.10 per million input tokens and $0.50 per million output tokens, aimed at focused, high-volume work, and it is the first GPT-6 model OpenAI is making available to Free and Go users in the ChatGPT desktop app as well as to paid accounts in ChatGPT Work, Codex, and the API.

Anthropic's release followed a different shape but a similar logic. Claude Opus 5.5 is the first model in a new Claude 5.5 family, and Anthropic is positioning it as its most advanced model to date while simultaneously making the case that it is the cheaper way to get top-tier work done. The company says Opus 5.5 performs at the level of Claude Fable 5.1 for most tasks and costs about 40% less to run than Opus 5 on typical workloads at default settings, with pricing at $4 per million input tokens and $20 per million output tokens — 20% cheaper than Opus 5 on input and output — and cached reads at $0.20 per million, which is 60% cheaper than the comparable Opus 5 cache rate. The context window is 1 million tokens with multimodal input, and Anthropic says the model leads on its own benchmarks in agentic coding, computer use, and knowledge work.

The '40% less to run' claim, read carefully

The "40% less to run" figure is the one to read carefully, because it is a typical-workload claim rather than a raw per-token price cut. On raw per-token pricing, GPT-6 Sol and Claude Opus 5.5 both landed cheaper than the tiers they relate to — Sol on raw per-token price, Opus 5.5 on workload cost — but the larger cost reduction Anthropic is leading with comes from the combination of cheaper per-token rates, cheaper cache reads, and — the part Anthropic is betting on — a model that spends fewer tokens and steps to finish the same work. That is a different kind of efficiency story than "we lowered the list price," and it is the kind of claim that only really lands when developers run the same workload on both models and compare the bills.

Distribution is where the same-day coincidence gets interesting. GPT-6 Sol and Luna are available in ChatGPT Work and Codex for most paid accounts and in the ChatGPT API, while Luna is also in the ChatGPT desktop app and for Free and Go users. Claude Opus 5.5 landed in GitHub Copilot the same day — for Pro+, Max, Business, and Enterprise subscribers in a phased rollout with no universal availability date yet — and Anthropic published the announcement at 16:31 UTC, the same minute it went live on anthropic.com. Putting a new flagship into Copilot the same day it launches is a distribution signal: Copilot is where a large share of professional developers already live, and a model that is inside the tool they are already using is a model they can try without changing their workflow.

What the timing says about the frontier

The shared theme is not that the frontier got cheaper in some abstract sense. It is that both labs chose the same day to make their newest capabilities easier to reach — OpenAI by building a three-tier GPT-6 stack and pushing the low end into free accounts, Anthropic by taking its most expensive tier and making it cheaper to run than the model it replaced, then putting it inside a developer tool developers were already using. The capability conversation that has dominated the last two weeks — the GPT-6 Astra "is this AGI-level?" debate — did not stop. It just stopped being the only conversation, which is itself a useful signal about where the news cycle is heading.

There is a subtle competitive read tucked inside the timing. Anthropic's Claude Fable 5.1 review from September 15 is the model Opus 5.5 is measured against — "performs at the level of Claude Fable 5.1 for most tasks" — which means Anthropic is effectively saying the new flagship absorbs the old premium tier's job and does it more cheaply, while Fable remains the cheaper option for users who do not need the full Opus ceiling. OpenAI, meanwhile, has built a three-tier GPT-6 stack and is pushing Luna into free accounts, which is a distribution play as much as a model play. The two labs are not copying each other, but they are converging on the same underlying idea: the next frontier move is as much about the shape of access and the total cost of doing real work as it is about raw capability.

If you are choosing between GPT-6 Sol and Claude Opus 5.5 for a multistep coding or knowledge-work workload in September 2026, the honest answer is that you are choosing between two reasonably priced ways to get flagship-adjacent performance, and the decision comes down to which ecosystem you are already in. Sol is the OpenAI middle rung — strong on coding per third-party benchmarks, available in ChatGPT Work, Codex, and the API, with a 1,050,000-token context window. Opus 5.5 is the Anthropic flagship replacement — strong on agentic coding, computer use, and knowledge work per Anthropic's own benchmarks, at a typical-workload cost about 40% below Opus 5, and now inside GitHub Copilot. The cheapest option on paper is Luna at $0.10 in and $0.50 out, but Luna is the focused high-volume tier, not the model you pick for a complex multistep agent.

The bigger read is that September 2026 is shaping up to be the month the frontier got cheaper on purpose. OpenAI expanded downward. Anthropic cut the flagship's effective cost. And both did it while the GPT-6 Astra conversation — the "is this AGI-level?" debate that has dominated the last two weeks — was still live. The task for anyone paying attention is not to decide which lab "won" the day, because neither did — both moved the same direction on the same afternoon, and the direction is the story. For the full pricing breakdown across all three models, see our GPT-6 Sol vs GPT-6 Luna vs Claude Opus 5.5 breakdown.

— Sources: OpenAI's GPT-6 Sol and Luna announcement (openai.com, Sep 22 2026); Anthropic's Claude Opus 5.5 announcement (anthropic.com/claude-opus-5-5, Sep 22 2026); OpenAI Developer Community announcement; GitHub Copilot changelog (github.blog, Sep 22 2026); TechCrunch; Artificial Analysis; LLM Stats; BenchLM.ai; OpenRouter; 9to5Mac; BNN Bloomberg; Yahoo Finance; Eigent AI; Coursiv; The Rundown; Kingy AI; CellCog; The New Stack.