Claude Haiku 5.5 vs GPT-6 Luna: Two Budget Bets on Oct 7

Two launches, one day

October 7, 2026 produced two releases aimed at the cheap end of the AI market, but they were aimed at very different audiences. OpenAI began rolling out GPT-6 in ChatGPT alongside a new capability called Intelligent UI. Anthropic shipped Claude Haiku 5.5, its smallest and least expensive model. One release tracker logged both on the same date. It listed Haiku 5.5 as its most recent entry among 140 releases from six providers, and placed OpenAI's GPT-6 rollout beside it, a day after Google's Gemini Nano Banana 2.1 5.

Comparing them is tempting, especially OpenAI's entry-tier GPT-6 Luna against Haiku 5.5. A close look shows they are not the same kind of product, and that distinction is the main takeaway.

What OpenAI shipped

GPT-6 comes in tiers. GPT-6 Sol is for Plus, Pro, Business and Enterprise subscribers, and GPT-6 Luna serves Free and Go users 1. Paid plans received it on October 7, and the free and Go tiers followed on October 8 1. OpenAI's announcement said the rollout to paid tiers would start immediately 2.

The main feature is Intelligent UI. Instead of returning only text, ChatGPT can respond with a purpose-built interface such as a clickable bike diagram, a savings calculator, a bill splitter or a map of road-trip stops 1. OpenAI says the interface is assembled progressively. A library of native components and a compiler render it while the model is still generating, and GPT-6 can start answering while it continues to reason 2.

OpenAI's speed claim is narrower than it first looks. GPT-6 Instant starts answering web-search questions 44% sooner on average than GPT-5.6 Instant, and OpenAI states plainly that this measures time to the first response, not total completion time 21. The feature applies to the Chat experience on web and mobile 1. Reaction on the developer forum was mixed. One user asked for an update to ChatGPT Classic, and another said the launch did nothing for developers 2.

What Anthropic shipped

Haiku 5.5 is a developer product, and it competes mainly on price. It costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, one-tenth of Haiku 4.5's $1 and $5 34. It has a 1M-token context window, up to 128K tokens of output, and adjustable effort 4. It is the first Haiku with an effort setting, and it closes out Anthropic's 5.5 family 5.

The sources agree on the list price but frame the savings differently. DataCamp notes that Anthropic claims an average saving of 75%, not 90% 3. The likely explanation is in Beam's breakdown. Above 100,000 tokens, every rate rises fivefold, to $0.50 input and $2.50 output 4. Long-context workloads therefore lose much of the discount, and teams that feed large documents to the model should budget for the higher rate.

The capability gains appear real. DataCamp reports that Haiku 5.5's OSWorld score rose from 15.7% to 72.4%. In its own hands-on test, both Haiku generations scored 24/24, but the Haiku 5.5 run cost $0.0098 compared with $0.25 3. Beam focuses on agent workflows, arguing that Haiku 5.5 makes subagents the cheap part of a pipeline. Its pricing table puts Sonnet 5.5 at $2/$10 and Opus 5.5 at $4/$20 4. The tracker likewise describes Haiku 5.5 as a low-cost helper agent alongside its larger siblings, aimed at summaries, classification and live customer support 5.

Why a price-matched comparison doesn't hold

The obvious matchup is the two "budget" models released the same day. None of the available reporting gives per-token pricing for GPT-6 Luna. It is described only as the model behind ChatGPT's Free and Go plans 1. Without that figure, any claim that Luna and Haiku 5.5 are price-matched would be speculation.

The more useful reading is that the two companies are competing in different places. OpenAI's October 7 launch is a consumer-interface bet. Its emphasis is on how answers look and how quickly they begin, and the cheap tier is a way to bring free users into that experience. Anthropic's launch is an infrastructure bet. It is selling volume work and agent orchestration at a price low enough that running many small calls becomes the default design.

A release tracker that lists both on the same line is accurate about the dates but can suggest a rivalry that does not quite exist. For developers choosing a model to build on, Haiku 5.5's published rates are concrete. Its main caveat is the 100K-token threshold. For people choosing a chatbot, the relevant question is whether Luna's interactive answers are useful, and cost per token does not factor in. Until OpenAI publishes API pricing for Luna, a direct cost comparison between the two is not possible.