OpenAI API Pricing 2026: GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna Cost Per Token Breakdown

As of October 30, 2026, GPT-6 Astra, OpenAI's flagship since September 3, 2026, costs $10.00 per million input tokens and $50.00 per million output. GPT-6.1 Sol, released September 29, costs $2.00/$10.00, GPT-6.1 Terra, released October 26, costs $1.00/$5.00, and GPT-6 Luna is the cheapest model not on a deprecation list at $0.10/$0.50. o1, o3-mini and o4-mini were shut down on October 23, and GPT-4o, still $2.50/$10.00, follows on January 14, 2027. That's the short answer. The longer answer — what a real task actually costs, and which models are about to disappear — is more interesting.

Headline per-token rates are the number everyone quotes and the number that matters least. What you actually pay depends on which model you route to, how many reasoning tokens it burns invisibly, whether your input is cached, and whether you run jobs in real time or in batch. Below is the full October 2026 price list, then the math that turns those rates into a monthly bill, plus the shutdown calendar: o1, o3-mini and o4-mini are already gone (October 23), and production traffic still pointed at o3 or GPT-5 (December 11) or GPT-4o (January 14) will break if nobody migrates it in time.

What Changed in OpenAI API Pricing (Updated October 30, 2026)

I re-checked OpenAI's pricing page, changelog, deprecations page and two API guides on October 30, 2026. Since my October 15 update, OpenAI released GPT-6.1 Terra, shut down o1, o3-mini and o4-mini, added 12-hour prompt caching, and released GPT-6 Rosalind to trusted-access organizations only. None of the standard rates in the main table below moved. The bigger story is the shutdown calendar: every GPT-4o, o3 and GPT-5 workload now has a date attached, and for some of them the recommended replacement costs more per token.

Sources, all opened October 30, 2026: OpenAI API pricing; OpenAI API changelog; OpenAI deprecations page; OpenAI reasoning guide; OpenAI prompt caching guide.

What the Recommended Replacements Cost at List Price

OpenAI's deprecations page names a replacement for every retiring model. Comparing list prices at identical token counts, the replacements land on both sides of the old rate: GPT-4o and o1 workloads get cheaper, while o3, GPT-5 and the mini and nano tiers move up. A multiple below 1.0x means the replacement is cheaper.

Rates and replacements from the OpenAI pricing and deprecations pages, opened October 30, 2026; multiples are my arithmetic on list prices. GPT-5.6 Sol uses its promotional $4.00/$20.00 rate, available at least through November 21, 2026. If Sol returned to the pre-cut rate that OpenAI's August 21 percentages imply ($5.00/$30.00), the Sol multiples would rise another 25% on input and 50% on output. Reasoning models can also use more or fewer tokens than their predecessors, so these ratios are a starting point for a test, not a forecast.

My Take, by Use Case

If you are still on GPT-4o or GPT-4o mini, this is the rare forced migration where the list price goes down: 25x on GPT-4o input and 1.5x on GPT-4o mini input. The work there is evaluating quality, not defending a budget. If you are on o3, GPT-5, GPT-5 mini or GPT-5 nano, the recommended replacements cost 2x to 8x more per token at identical token counts, so price that migration before December 11 instead of discovering it on the invoice. And if you run reasoning workloads, test GPT-6.1 Terra at $1.00/$5.00 before defaulting to Sol; GPT-6.1 Sol at $2.00 input is also half the promotional GPT-5.6 Sol rate of $4.00. Those are my judgment calls from list prices, not benchmark results.

OpenAI API Pricing 2026: The Full Per-Token Breakdown

OpenAI API pricing as of October 30, 2026 covers the models below across four generations: the GPT-6 line (Astra, GPT-6.1 Sol, GPT-6.1 Terra, Sol, Luna), the GPT-5.6 line (Sol, Terra, Luna), the GPT-5 and GPT-4o family, and the legacy o-series reasoning models. Input rates run from $0.10 to $10.00 per million tokens for the models listed, output from $0.50 to $50.00, and you pay only for tokens actually processed. Pro-class models such as o3-pro ($20.00/$80.00) and GPT-5.5 Pro ($30.00/$180.00) are priced higher and are not in the table.

Rates are per 1M tokens, USD, standard tier, as published on OpenAI's API pricing documentation, with release dates from OpenAI's API changelog and shutdown dates from OpenAI's deprecations page, opened October 30, 2026. OpenAI revises pricing frequently, and GPT-6 Astra launched at $10.00/$50.00, five times GPT-6.1 Sol's $2.00/$10.00. Always confirm against the live pricing page before committing a budget.

Batch, Fast and Ultrafast: How the Processing Tiers Change the Rate

The standard rate is only one of four tiers on OpenAI's pricing page. Batch (labeled Flex/Batch) runs at half the standard rate, Fast mode (renamed from Priority processing on July 30, 2026) doubles it, and Ultrafast, which OpenAI added for GPT-6.1 Sol on October 8, costs six times standard. Per million tokens, input / output:

Source: OpenAI API pricing, opened October 30, 2026. The pricing page lists Ultrafast only for Astra and GPT-6.1 Sol. Regional (data residency) endpoints are charged a 10% uplift for eligible models released on or after March 5, 2026. All rates here are for prompts up to 272K input tokens.

GPT-6 Astra: OpenAI's New $10/$50 Flagship, Launched September 3, 2026

OpenAI launched GPT-6 Astra on September 3, 2026, priced at $10.00 per million input tokens and $50.00 per million output — a new premium tier sitting above every existing model, including GPT-5.6 Sol. Cached input runs $1.00 per million (a 90% discount), Fast mode doubles the standard rate for lower latency, and Batch/Flex processing halves it to $5.00/$25.00, and Ultrafast mode, added September 29, lists at $60.00/$300.00. According to CNBC's September 3 report, Astra rolled out first to organizations in OpenAI's cybersecurity access program before expanding to the general API and Azure/AWS Bedrock.

Astra is not a Sol replacement — both remain live in the API at their own price points. This likely means OpenAI is now running a genuine multi-tier premium strategy rather than a single ever-cheaper flagship: Astra for the highest-stakes agentic and coding work, Sol for general frontier reasoning, and Terra/Luna for everyday and high-volume traffic. Whether that pricing gap holds is unproven — OpenAI cut Sol's price by 20% on input and 33% on output on August 21, which implies an earlier rate of about $5.00/$30.00, and GPT-6.1 Sol has since arrived at $2.00/$10.00, so a price move on Astra within its first few months would not be surprising.

GPT-5.6 Sol, Terra, and Luna: Why OpenAI's Flagship Got More Expensive, Then Cheaper Again

OpenAI released three GPT-5.6 tiers on July 9, 2026: Sol (flagship), Terra (mid-tier), and Luna (budget). Sol's launch price was above GPT-5's $1.25 per million input tokens; OpenAI's later cut of 20% on input implies a launch rate of about $5.00. According to OpenAI's own GPT-5.6 announcement, the reason was infrastructure: Sol is the first OpenAI model to run in production on dedicated Cerebras wafer-scale chips, delivering up to 750 tokens per second versus roughly 70 tokens per second on a typical Nvidia H100 cluster running the same weights.

On August 21, 2026, OpenAI cut Sol's API price by 20% on input and 33% on output — to $4.00 and $20.00 per million tokens — with that promotional rate available at least through November 21, 2026. Terra and Luna, cut 20% and 80% respectively on July 30, sit at $2.00 and $0.20 input and have not moved since, per the changelog. In my read, most existing production traffic can migrate to one of those two tiers without much dollar movement, leaving Sol (and now Astra) for the narrower slice of latency-critical, highest-capability workloads.

Since then, the GPT-6 line has undercut the GPT-5.6 tiers. OpenAI released GPT-6 Sol and GPT-6 Luna on September 22 at $2.00/$10.00 and $0.10/$0.50, then GPT-6.1 Sol on September 29 at the same $2.00/$10.00 with cached input at $0.10, and GPT-6.1 Terra on October 26 at $1.00/$5.00, half GPT-5.6 Terra's input price and about 58% below its output price, per OpenAI's API changelog. For teams building agentic systems, OpenAI also offers the agents API as an alternative pricing model optimized for autonomous workflows.

What OpenAI API Pricing Looks Like Per Real Task

A token is roughly 0.75 words, so 1,000 tokens is about 750 words. Most real requests are small — a few hundred tokens in, a few hundred out. The table below prices six common workloads on the model you'd actually pick, assuming standard (uncached) input.

*The o3 example assumes about 12K output tokens, mostly hidden reasoning, billed at the output rate. The 12K is my illustrative assumption, not an OpenAI figure; OpenAI's reasoning guide says reasoning tokens are billed as output tokens.

The lesson: individual calls are cheap, but volume compounds fast. A product doing 5 million GPT-4o summaries a month at $0.026 each is spending $130,000 — and much of that is avoidable with the right routing and caching, covered below. The same 5 million summaries on GPT-6 Luna at about $0.0011 each is roughly $5,500 a month at list price, assuming Luna produces comparable output and token counts. For how this rolls up across the industry, see our AI Spending dashboard.

The Same Tasks After the Forced Migration

With GPT-4o and GPT-4o mini shutting down January 14, 2027 and GPT-6 Luna named as the replacement, here is the same token math re-run at list prices, keeping the token counts identical. The o3 row uses GPT-5.6 Sol at its promotional rate.

List prices from OpenAI's pricing page, opened October 30, 2026; arithmetic is mine. Real bills depend on how many tokens each model actually generates for your prompts, which can differ from the old model's, especially for reasoning tasks.

Why Reasoning Models Cost More Per Task: o1 Is Gone and o3 Follows December 11

Reasoning models think before they answer, and that thinking is made of tokens you pay for. OpenAI's reasoning guide says reasoning tokens are not visible through the API, still occupy space in the model's context window, and are billed as output tokens. At o3's $8.00-per-million output rate, an answer that generates 12,000 output tokens, my illustrative figure, costs $0.096 in output alone, against $0.005 for a 500-token GPT-4o answer at $10.00. That is why o3, with a lower headline price than GPT-4o ($2.00/$8.00 versus $2.50/$10.00), can cost more than ten times as much for the same question.

The retirements are now underway. OpenAI's deprecations page says access to o1, o1-pro, o3-mini and o4-mini was shut down on October 23, 2026, although the pricing page still lists their rates. Full o3 and o3-pro follow on December 11, 2026 alongside the original GPT-5 snapshot family. OpenAI names GPT-5.6 Sol as the replacement for o1, o3-mini, o3 and their pro variants, and GPT-5.6 Terra for o4-mini.

The practical rule: don't send a task to Sol or Astra unless GPT-6 Luna or Terra demonstrably fails it. Reasoning and flagship-tier models are a scalpel, not a default. For simple classification, routing and extraction, start on GPT-6 Luna and move up only when it fails your own evaluation set.

OpenAI API Pricing vs Anthropic and Google in 2026

OpenAI isn't priced in a vacuum. Anthropic's Claude pricing and Google's Gemini API pricing set the competitive floor, and GPT-6 Astra now ties Anthropic's Claude Fable 5.1 as the most expensive model listed below. Here's how the top and budget tiers line up: OpenAI's rates were re-checked October 30, 2026, while the Anthropic and Google rows are as of October 15, 2026 and were not re-verified in this update.

Sources: OpenAI, Anthropic and Google API pricing documentation; OpenAI accessed October 30, 2026, Anthropic and Google October 15, 2026. Claude Haiku 5.5's rate applies to prompts up to 100,000 tokens. Gemini 4 Argon's rate runs through November 30, 2026, then rises to $4.00/$20.00; Gemini 3.8 Flash's runs through December 31, 2026, then rises to $1.50/$7.50.

The takeaway: at the top, GPT-6 Astra and Claude Fable 5.1 both list at $10/$50, 2.5x Claude Opus 5.5 and 5x Gemini 4 Argon's introductory rate on input. In the middle, GPT-6.1 Sol and Claude Sonnet 5.5 are priced identically at $2/$10. At the budget end, GPT-6 Luna and Claude Haiku 5.5 both list at $0.10/$0.50, two-thirds of GPT-4o mini's input price. Google's introductory rates are the ones to watch: Argon's doubles on December 1 and Gemini 3.8 Flash's on January 1, 2027. For the full competitive picture, see our OpenAI vs Anthropic enterprise breakdown.

Four Levers That Cut an OpenAI API Bill: Caching, Batch, Routing and Prompt Trimming

Prompt caching is usually the biggest of the four, because most production apps resend the same instructions on every request. OpenAI's prompt caching guide says cached input can be discounted by up to 95%. For GPT-5.6 and later, a 30-minute cache write costs 1.25x the standard input rate, a 12-hour write in the Responses API costs 2.5x, and reads cost 0.1x on most models and 0.05x on GPT-6.1 Sol and GPT-6.1 Terra. The minimum cacheable prompt is 1,024 tokens. At the 0.1x read rate, a 30-minute write pays for itself on the second request, 1.35x versus 2x uncached, and a 12-hour write on the third, 2.7x versus 3x.

Here is the arithmetic on a hypothetical workload, at list prices and not a forecast. Two million calls a month, each carrying the same 15,000-token system prompt, is 30 billion input tokens. On GPT-6.1 Sol at $2.00 per million, that is $60,000 a month uncached. If 95% of those tokens are cache reads at $0.10 per million and 5% are 30-minute writes at $2.50, the input bill is $2,850 plus $3,750, about $6,600, roughly 89% lower. Output tokens are unaffected by caching, and your real hit rate depends on how often identical prefixes arrive inside the cache lifetime.

What the headline pricing misses

The clean story — "OpenAI keeps getting cheaper" — doesn't survive contact with GPT-6 Astra. Astra costs 2.5x GPT-5.6 Sol's promotional rate and 50x GPT-5.6 Luna on input. Anyone budgeting off last year's per-token math and routing production traffic straight to the newest flagship without checking current rates first could see a bill spike, not a discount. Below the flagship the pattern is different: GPT-6.1 Sol costs half of GPT-5.6 Sol's promotional input rate, and GPT-6.1 Terra costs half of GPT-5.6 Terra's.

A second caveat: the discount math above assumes standard list pricing. Enterprise customers frequently negotiate committed-use or volume contracts off-list, and those terms aren't published, so a large team's actual per-token cost can differ meaningfully from every number in this article. This likely means the gap between GPT-6 Astra and its competitors is smaller in practice for high-volume enterprise buyers than the public price list suggests, though there's no public data to size that gap precisely. A third: Sol's $4.00/$20.00 rate is a promotion OpenAI says is available at least through November 21, 2026 — anyone building a 2027 budget off today's promotional price is guessing, not calculating. A fourth: every rate in this article is standard processing for prompts up to 272K input tokens. OpenAI lists separate rates for longer prompts, which I did not capture, and regional data-residency endpoints carry a 10% uplift on eligible models released on or after March 5, 2026. Cross-model tracking sites such as Artificial Analysis are a reasonable way to sanity-check list prices against real-world benchmarking data before committing a budget.