AI API pricing this week: GPT-6 Cyber at $20/$100, Gemini Omni 1.2 Flash at $0.07 a second and Opus 4.8 fast mode ends

38 minutes ago

9 min read

AI API pricing this week cover: GPT-6 Cyber at $20/$100, Gemini Omni 1.2 Flash at $0.07 a second and Opus 4.8 fast mode ends. Data Studios original cover

The week of November 2 to November 6, 2026 brought five official API price and terms changes from four providers. OpenAI opened GPT-6 Cyber to approved Daybreak Red organizations at $20 per 1M input tokens and $100 per 1M output tokens, and priced its new GPT-Image-2.6 Sunburst and Flare image models at $30 per 1M image output tokens with no Batch rate on the pricing page. Google released Gemini Omni 1.2 Flash at $0.07 per second of 720p video, which the company describes as 30% below Gemini Omni 1.1 Flash.

On the Anthropic side, the Claude API set November 30, 2026 as the removal date for fast mode on Claude Opus 4.8, where it cost $10 input and $50 output per 1M tokens, and points developers to fast mode on Claude Opus 5.5 at $8 and $40. xAI retired the grok-imagine-image-quality slug on November 2 and now serves those calls with grok-imagine-image-2.0 at low quality, which the documentation says costs $0.01 less per image at every resolution.

All prices below are in USD per 1M tokens unless the provider uses another unit, with input and output listed separately and cache and batch rates shown where they exist. Prices as listed on the official pricing pages and documentation of each provider on November 6, 2026.

··········

ALL API PRICE AND TERMS CHANGES THIS WEEK.

The table lists every verified change. One row is one priced item, so a single launch can occupy several rows.

........

Provider	Model	Item	Before	After	Effective date
OpenAI	GPT-6 Cyber	Input / cached input	New model	$20.00 / $2.00	Nov 5, 2026
OpenAI	GPT-6 Cyber	Output	New model	$100.00	Nov 5, 2026
OpenAI	GPT-6 Cyber	Cache write 30 min / 12 h	New model	$25.00 / $50.00	Nov 5, 2026
OpenAI	GPT-Image-2.6 Sunburst and Flare	Image output	GPT-Image-2.5 Batch: $15.00	$30.00 standard, no Batch rate listed	Nov 4, 2026
OpenAI	GPT-Image-2.6 Sunburst and Flare	Text input / image input	GPT-Image-2.5 Batch: $2.50 / $4.00	$5.00 / $8.00 standard	Nov 4, 2026
OpenAI	GPT-Image-2.6 Sunburst and Flare	Cached text / cached image input	GPT-Image-2.5 Batch: $0.625 / $1.00	$1.25 / $2.00 standard	Nov 4, 2026
Google	Gemini Omni 1.2 Flash	720p video output	About $0.10 per second (Omni 1.1 Flash)	$0.07 per second	Nov 5, 2026
Anthropic	Claude Opus 4.8	Fast mode input / output	$10.00 / $50.00	Removed, requests return an error	Nov 30, 2026 (deprecated Oct 30)
Anthropic	Claude Opus 5.5	Fast mode input / output (migration target)	Not applicable	$8.00 / $40.00	Available now
xAI	grok-imagine-image-quality	Per image, 1K / 2K	$0.01 above the new rate	$0.04 / $0.06 (grok-imagine-image-2.0, low)	Nov 2, 2026
xAI	grok-imagine-image-quality	Edits, per input image	Not stated	$0.01	Nov 2, 2026

........

The official changelogs of DeepSeek and Mistral showed no entries for the week, and the Claude API release notes for November 2 to 5 added features (message threads in beta, a text watermark, IP allowlisting in beta) without price changes.

··········

OPENAI GPT-6 CYBER API PRICING FOR DAYBREAK RED ORGANIZATIONS.

OpenAI added GPT-6 Cyber to its API changelog on November 5 as a model for investigating vulnerabilities, reproducing them in an authorized test environment and verifying fixes. Access depends on Daybreak Red approval for the organization, and the model page tells developers to complete the Daybreak setup and confirm access with their account team before the first request. Our launch coverage, OpenAI releases GPT-6 Cyber for Daybreak Red at $20 per 1M input tokens, describes the access program in detail.

The pricing page lists GPT-6 Cyber at $20.00 input, $2.00 cached input and $100.00 output per 1M tokens, with cache writes at $25.00 for a 30-minute cache and $50.00 for a 12-hour cache. No Batch rate is listed for the model. Against GPT-6.1 Sol, OpenAI's general frontier model at $2.00 input and $10.00 output, the cyber model costs ten times as much per token in both directions. In the agent session described in the workload section below (50 turns, 40,000 input tokens per turn of which 35,000 are cached, 1,500 output tokens per turn), GPT-6 Cyber comes to $16.00 per session against $1.43 on GPT-6.1 Sol.

Our launch article compared GPT-6 Cyber with GPT-5.6-Cyber at $12.50 input and $75.00 output. The official pricing page no longer lists GPT-5.6-Cyber, so that comparison could not be re-verified on November 6 and is left out of the table above. For a cross-vendor view of the restricted cyber models, see our GPT-6 Cyber vs Gemini 4 Carbon comparison.

··········

GPT-IMAGE-2.6 SUNBURST AND FLARE PRICING AND THE MISSING BATCH RATE.

OpenAI released gpt-image-2.6-sunburst and gpt-image-2.6-flare on November 4 for the v1/images/generations and v1/images/edits endpoints, with adjustable complexity, SVG generation and animated GIF output. The standard prices per 1M tokens are $5.00 for text input, $1.25 for cached text input, $8.00 for image input, $2.00 for cached image input and $30.00 for image output, identical for both variants.

The change for buyers sits in the second pricing table. GPT-Image-2.5 Sunburst and Flare carry Batch rates of $4.00 image input, $1.00 cached image input, $2.50 text input and $15.00 image output, half the standard prices. The pricing page lists no Batch rates for the 2.6 models. A workload that generates 100 million image output tokens through Batch on GPT-Image-2.5 costs $1,500 in output; the same volume on GPT-Image-2.6 at the only listed rate costs $3,000. The model page does include v1/batch in its list of supported endpoints without giving a Batch price, so teams that rely on Batch should test billing on a small job before moving volume.

Our launch article, OpenAI prices GPT-Image-2.6 at $30 per 1M image tokens with no Batch discount, and the GPT-Image-2.6 vs GPT-Image-2.5 comparison cover the feature differences.

··········

GEMINI OMNI 1.2 FLASH VIDEO PRICING AT $0.07 PER SECOND.

Google announced Gemini Omni 1.2 Flash on November 5 and the Google Cloud blog marked it generally available on November 6. Both posts state a price of $0.07 per second of 720p video and a 30% reduction from Gemini Omni 1.1 Flash, with similar savings at other resolutions that Google does not quantify. The model is generally available in the Gemini API, can be tried in Google AI Studio and the Google Cloud console, supports Provisioned Throughput, and produces clips of up to 15 seconds.

The Gemini API pricing page, as fetched on November 6, still lists only gemini-omni-1.1-flash: $1.50 per 1M input tokens, $9.00 per 1M text output tokens and $17.50 per 1M video output tokens, with 720p video billed at 5,792 output tokens per second, which the page rounds to approximately $0.10 per second. The exact figure is $0.1014 per second, so the move to $0.07 is a 30.9% cut. Token-level rates for Omni 1.2 Flash were not on that page on November 6. A batch of 1,000 clips of 15 seconds at 720p costs $1,050 on Omni 1.2 Flash against about $1,520 on Omni 1.1 Flash.

Full details are in Google prices Gemini Omni 1.2 Flash at $0.07 per second, 30% below Omni 1.1.

··········

CLAUDE OPUS 4.8 FAST MODE REMOVAL AND THE OPUS 5.5 FAST MODE PRICE.

The Claude API release notes of October 30 deprecated fast mode on Claude Opus 4.8, with removal on November 30, 2026. After that date, requests to claude-opus-4-8 with speed set to fast return an error, and Anthropic's documentation directs developers to fast mode on Claude Opus 5.5. Fast mode on Opus 4.8 and Opus 5 is a research preview that needs the fast-mode-2026-02-01 beta header, while on Opus 5.5 it is generally available.

Fast mode on Opus 4.8 costs $10 input and $50 output per 1M tokens; on Opus 5.5 it costs $8 and $40, 20% less, against standard Opus 5.5 prices of $4 and $20. The fast mode rates apply across the full context window, prompt caching multipliers apply on top, and fast mode is not available through the Batch API. Since Opus 5.5 bills cache hits at 0.05 times base input and Opus 4.8 at 0.1 times, the cached share of an agent loop also gets cheaper: the reference agent session costs $8.00 on Opus 4.8 fast mode and $5.70 on Opus 5.5 fast mode, a 28.75% reduction, while the same session in standard mode on Opus 5.5 costs $2.85.

The deprecation is covered in Claude API deprecates Opus 4.8 fast mode ahead of November 30 removal.

··········

XAI RETIRES GROK-IMAGINE-IMAGE-QUALITY AND BILLS THE GROK-IMAGINE-IMAGE-2.0 LOW RATE.

xAI retired the grok-imagine-image-quality slug on November 2, 2026, after a 60-day notice period that began on September 2. Requests to that slug and to grok-imagine-image-pro, which already redirected to it, are now served on v1/images/generations and v1/images/edits by grok-imagine-image-2.0 with quality set to low, with no code change required. grok-imagine-image (version 1.0) is unaffected.

The migration guide prices the low setting at $0.04 per 1K image and $0.06 per 2K image, plus $0.01 per input image on edits, and states that this is $0.01 per image cheaper than grok-imagine-image-quality at every resolution. xAI does not print the old rate on that page, so the implied previous prices of $0.05 and $0.07 are derived from that statement. A job of 10,000 images costs $400 at 1K and $600 at 2K at the new rate, $100 less in each case. Billing follows the quality actually served: with quality set to auto, generation runs at low and editing at medium, and the migration page gives no price for medium. Our Grok Imagine guide to generation, paid access and API pricing covers the rest of the Imagine API.

··········

WHAT THREE TYPICAL WORKLOADS NOW COST ACROSS THE MAIN MODELS.

The comparison uses the standard list prices on each provider's official pricing page on November 6, 2026, with three workloads. A customer-support chatbot handles 10,000 requests a day for 30 days, each with 3,000 input tokens, of which 2,000 are a reused system prompt billed at the cached-input rate, and 400 output tokens. A coding agent session runs 50 turns of 40,000 input tokens, 35,000 of them cached, and 1,500 output tokens per turn. A long-document job summarizes 1,000 documents of 90,000 input tokens into 2,000 output tokens each through the Batch API, without caching.

Cache write surcharges and Gemini cache storage fees are excluded, because with thousands of requests a day reusing the same prefix they are a small share of the bill. Gemini 4 Argon appears twice because its introductory prices of $2 and $10 end on November 30 and the $4 and $20 rates start on December 1; Gemini 3.8 Flash is shown at its introductory $0.75 and $3.75, which run through December 31 before $1.50 and $7.50 apply on January 1, 2027. Grok 4.7 is left out because the xAI models page lists no cached-input or Batch rate for it.

........

Model	Support chatbot, per month	Coding agent, per session	Coding agent, 1,000 sessions	1,000 documents, Batch
GPT-6.1 Sol	$1,860.00	$1.43	$1,425.00	$100.00
GPT-6.1 Terra	$930.00	$0.71	$712.50	$50.00
GPT-6 Luna	$96.00	$0.08	$80.00	$5.00
Claude Opus 5.5	$3,720.00	$2.85	$2,850.00	$200.00
Claude Sonnet 5.5	$1,860.00	$1.43	$1,425.00	$100.00
Claude Haiku 5.5	$96.00	$0.08	$80.00	$5.00
Gemini 4 Argon, to Nov 30	$1,860.00	$1.43	$1,425.00	$100.00
Gemini 4 Argon, from Dec 1	$3,720.00	$2.85	$2,850.00	$200.00
Gemini 3.8 Flash, to Dec 31	$720.00	$0.60	$600.00	$37.50

........

The arithmetic for the chatbot is 300,000 requests multiplied by (1,000 uncached input tokens at the input rate, 2,000 cached tokens at the cached rate and 400 output tokens at the output rate). On GPT-6.1 Terra that is 300,000 x (1,000 x $1.00 + 2,000 x $0.05 + 400 x $5.00) / 1,000,000 = $930. Claude Haiku 5.5 is priced at $0.10 and $0.50 only for prompts up to 100,000 tokens, so the 90,000-token documents stay in the lower tier; above that threshold its Batch rates rise to $0.25 and $1.25.

At the frontier tier, GPT-6.1 Sol, Claude Sonnet 5.5 and Gemini 4 Argon at its introductory price cost the same on all three workloads because their list prices coincide at $2 and $10 with $0.10 cache hits. From December 1, Gemini 4 Argon moves to the Claude Opus 5.5 price level and doubles each figure. Gemini 3.8 Flash at its introductory price is the cheapest of the mid-tier options in the table, ahead of GPT-6.1 Terra by about 23% on the chatbot, while GPT-6 Luna and Claude Haiku 5.5 share identical list prices at the low end.

··········

AI API PRICING IN THE WEEK OF NOVEMBER 2: CONFIRMED FACTS AND OPEN QUESTIONS.

The confirmed changes split into two directions. Video and image generation got cheaper per unit, with Gemini Omni 1.2 Flash at $0.07 per second of 720p video and xAI's retired quality slug now billed $0.01 less per image. Specialized and newer models went the other way for some buyers: GPT-6 Cyber sits at ten times the per-token price of GPT-6.1 Sol, and GPT-Image-2.6 has no listed Batch rate, which doubles the effective output price for workloads that ran GPT-Image-2.5 through Batch. The Claude Opus 4.8 fast mode removal on November 30 forces a migration that lowers the per-token fast mode price by 20% and the reference agent session by 28.75%.

Several points remain open on the official pages. OpenAI lists v1/batch as a supported endpoint for GPT-Image-2.6 without publishing a Batch price. The Gemini API pricing page did not yet carry token-level rates for Omni 1.2 Flash on November 6, so only the per-second figure from Google's announcements is confirmed. xAI has not published a price for grok-imagine-image-2.0 at medium quality on its migration page, which matters for edits made with quality set to auto. Two deadlines also fall in the coming weeks: the end of Gemini 4 Argon introductory pricing on November 30 and the Opus 4.8 fast mode removal on the same day.

··········

·····

DATA STUDIOS

·····

datastudios.org