Frontier Log

AI Lab Speed Tiers, Compared

OpenAI, Anthropic and Google all take 50% off for batch jobs. OpenAI's Ultrafast costs 3x Standard on GPT-6.2 Astra, and 6x from Mar 1, 2027.

Column charts of output prices per 1M tokens by service tier, one panel per lab, each column's height its multiple of the lab's Standard price. OpenAI, GPT-6.2 Astra: Batch $20 (24 hours), 8-hour window $24, 2-hour window $28, Flex not offered on GPT-6.2 (was 50% off), Standard $40, Fast $80 (2x), Ultrafast $120 (3x) to Feb 28, 2027 and $240 (6x) from Mar 1, 2027, drawn dashed. Anthropic, Claude Fable 5.2: Batch $20, Standard $40, fast mode $80 (2x). Google, Gemini 4 Carbon: Batch $5, Flex $5, Standard $10, Priority $18 (1.8x).
Output prices per 1M tokens on each lab's pricing page, Nov 20, 2026. All three labs charge half of Standard for batch jobs. OpenAI's Ultrafast costs 3x Standard on GPT-6.2 Astra until Feb 28, 2027, and 6x from Mar 1. OpenAI's window prices are our arithmetic. Download

OpenAI added two cheaper ways to wait for an answer on Nov 17, 2026. Its API changelog says model token rates are 30% lower than Standard for a two-hour background completion window. They are 40% lower for an eight-hour window, and both windows are in public beta. Its background mode guide adds that new models, starting with GPT-6.2 Astra and Sol, launch without Flex processing.

Waiting still pays most on Batch

OpenAI's Batch API has run since Apr 15, 2024, by its changelog. Its Batch guide promises results within 24 hours at 50% lower costs. The pricing page lists Batch output for GPT-6.2 Astra at $20 per 1M tokens, half of the $40 Standard price.

The new windows sit between the two, and the API targets completion within the window a request picks. On GPT-6.2 Astra, 40% off Standard comes to $24 for the eight-hour window and 30% off comes to $28 for the two-hour window. The windows work on GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna, and they share the Batch rate limit with Batch API jobs.

Flex charges Batch API rates for slower responses, the Flex guide says. Flex is unavailable on GPT-6.2 Astra and Sol, and OpenAI won't add it to future models. OpenAI is winding down the beta and will share a retirement date later. GPT-6 Astra still lists Flex output at $25 per 1M tokens on the pricing page.

Speed costs up to 6x at OpenAI

OpenAI launched Priority processing on Jun 27, 2025, by its changelog. The pricing page says Priority processing was renamed Fast mode on Jul 30, 2026. Fast mode costs twice the Standard rate, and the Fast mode guide promises up to 2.5x faster speeds. GPT-6.2 Astra's Fast output costs $80.

OpenAI announced Ultrafast on Aug 13, 2026, as a new tier for GPT-5.6 Sol in limited preview. The changelog says it runs up to 14x faster than Standard processing. For GPT-6.2 Astra and Sol, the pricing page says Ultrafast costs 3 times the Standard rate through Feb 28, 2027, and 6 times from Mar 1, 2027. On GPT-6.2 Astra, output costs $120 until then and $240 after. The page calls the $240 rate a list rate. Ultrafast already costs 6 times Standard on GPT-6 Astra and GPT-6.1 Sol.

Anthropic and Google also halve batch prices

Anthropic's Message Batches API costs 50% less than standard prices. Most batches finish within an hour, and a batch expires if it doesn't finish within 24 hours. It opened on Oct 8, 2024, by Anthropic's release notes. Fast mode launched in research preview on Feb 7, 2026, and reached Claude Fable 5.2 on Nov 9, 2026. The pricing page lists Fable 5.2's fast mode output at $80, against $40 at Standard.

Google launched Batch Mode for the Gemini API on Jul 7, 2025, and added Flex and Priority on Apr 1, 2026, its changelog says. Its optimization guide lists Batch at 50% off within 24 hours, and Flex at 50% off with a 1 to 15 minute target. The same guide says Priority costs 75% to 100% more than Standard.

On Gemini 4 Carbon, the pricing page lists Priority output at $18, against $10 at Standard. Those prices hold through Nov 30, 2026. From Dec 1, every Gemini 4 Carbon tier doubles, so the multiples stay the same.

GPT-6.2 Astra output price

ModelValueSources
GPT-6.2 Astra and SolBatch$20developers.openai.comFlagship models, Speed: Batch, Short context, gpt-6.2-astra, Output: $20.00
GPT-6.2 Astra and Sol8-hour window$24developers.openai.comnote under the Flagship models table: 'model token rates for eligible background requests are ... 40% below Standard for the 8-hour completion window', applied to the Standard output price of $40.00 (our arithmetic)
GPT-6.2 Astra and Sol2-hour window$28developers.openai.comnote under the Flagship models table: '30% below Standard for the 2-hour completion window', applied to the Standard output price of $40.00 (our arithmetic)
GPT-6.2 AstraStandard$40Announcement'Pricing and availability', the 'Standard API rates per million tokens are ...' sentence
GPT-6.2 Astra and SolFast$80developers.openai.comFlagship models, Speed: Fast, Short context, gpt-6.2-astra, Output: $80.00
GPT-6.2 Astra and SolUltrafast, through Feb 28, 2027$120developers.openai.comFlagship models, Speed: Ultrafast, table under 'For GPT-6.2 Astra and GPT-6.2 Sol, Ultrafast costs 3 times the Standard rate through February 28, 2027', gpt-6.2-astra, Output: $120.00
GPT-6.2 Astra and SolUltrafast, from Mar 1, 2027$240developers.openai.comFlagship models, Speed: Ultrafast, 'GPT-6.2 list rates' table ('6 times the Standard rate starting March 1, 2027'), gpt-6.2-astra, Output: $240.00

Claude Fable 5.2 output price

ModelValueSources
Claude Fable 5.2 and Mythos 5.2Batch$20platform.claude.comBatch processing table, Claude Fable 5.2, Batch tokens Output: $20 / MTok
Claude Fable 5.2Standard$40Announcementparagraph beginning 'Testers have found that Fable 5.2 and Mythos 5.2 communicate more clearly', the price at its end
Claude Fable 5.2 and Mythos 5.2Fast mode$80platform.claude.comFast mode pricing table, Claude Fable 5.2, Output: $80 USD / MTok

Gemini 4 Carbon output price

ModelValueSources
Gemini 4 CarbonBatch$5ai.google.devGemini 4 Carbon, Batch tab, Paid Tier Output price: $5.00 through November 30, 2026
Gemini 4 CarbonFlex$5ai.google.devGemini 4 Carbon, Flex tab, Paid Tier Output price: $5.00 through November 30, 2026
Gemini 4 CarbonStandard$10Announcementparagraph on launch introductory price (same as Argon)
Gemini 4 CarbonPriority$18ai.google.devGemini 4 Carbon, Priority tab, Paid Tier Output price: $18.00 through November 30, 2026

Sources

Models

Related

Tags