GPT Image 2.6 Pricing: Official API Rates & Cost Guide

How Much Does GPT Image 2.6 Cost? Official $30/1M Image Output Rates

GPT Image 2.6 costs $5.00 per million text input tokens, $8.00 per million image input tokens, and $30.00 per million image output tokens in the official OpenAI API. Standard input caching reduces text input costs to $1.25 per million tokens and image input costs to $2.00 per million tokens. The pricing structure applies identically across both released model variants: gpt-image-2.6-sunburst for high-fidelity composition and gpt-image-2.6-flare for low-latency generation. While raw token unit costs match the preceding GPT Image 2.5 generation, effective per-image inference budgets depend directly on image resolution, quality parameters, and output token usage.

Source: @ChrisGPT

Key Takeaways

• Identical Token Baseline: Official API token rates remain identical to GPT Image 2.5 at $5.00/1M text input, $8.00/1M image input, and $30.00/1M image output.
• Dual Model Portfolio: Both gpt-image-2.6-sunburst and gpt-image-2.6-flare share the exact same token pricing schedule.
• Cache Reductions: Cached prompt text costs $1.25 per million tokens, while cached image inputs drop to $2.00 per million tokens (a 75% reduction on repeated inputs).
• No Complexity Surcharge: The new complexity steering control (low, medium, high) carries no standalone platform API fee.
• Output Token Overhead: Early developer benchmarks show version 2.6 can consume 3.3x to 4.2x more output tokens than version 2.5 on certain low and medium quality calls, altering effective run budgets.
• No Batch Discounts: OpenAI launched GPT Image 2.6 with the Batch API (v1/batches) and Responses API marked as unsupported, removing the 50% discount path available on legacy image models.

GPT Image 2.6 Pricing Breakdown

OpenAI officially introduced GPT Image 2.6 on November 4, 2026, deploying the family through the standard Images API endpoints (v1/images/generations and v1/images/edits) as announced on the OpenAI official launch announcement.

Billing for GPT Image 2.6 is meter-based on processed tokens rather than fixed per-asset tiering. The model accepts multimodal inputs (text prompts and image reference files or transparent PNG markup overlays) and emits generated visual media including PNG, JPEG, WebP, SVG vectors, and animated GIFs.

The table below reflects official OpenAI API token rates as verified on November 5, 2026.

Billing Component	Official Price (USD / 1M Tokens)	Cached Rate (USD / 1M Tokens)	Applies To	Notes / Constraints
Text Input	$5.00	$1.25	Sunburst & Flare	Prompt instructions, negative guidance, edit descriptions
Image Input	$8.00	$2.00	Sunburst & Flare	Base images, reference character files, PNG markup layers
Image Output	$30.00	N/A	Sunburst & Flare	Standard generation, SVG code export, animated GIF frames
Complexity Setting	$0.00 (No extra fee)	N/A	Sunburst & Flare	Accepts low, medium, high (default high)
Batch API Discount	Not supported	N/A	Sunburst & Flare	v1/batches disabled at launch
Markup Overlay	Standard Image Input	$2.00	Sunburst & Flare	Replaces binary masks; mutually exclusive with mask

Both primary models operate under identical pricing bands. The distinction between them is operational: gpt-image-2.6-flare optimizes for iteration speed, capable of producing an image in as little as 6 seconds at low quality, whereas gpt-image-2.6-sunburst prioritizes prompt adherence, typographic accuracy, and structural consistency across successive edit iterations.

What That Costs in Practice

Because the API bills per token rather than per delivered image file, total request cost hinges on the input prompt payload, whether source images or markup layers are passed, and how many output tokens the diffusion or autoregressive synthesis process emits.

A standard single-image request at 1024x1024 resolution typically consumes roughly 800 to 1,000 output tokens depending on quality parameters (low, medium, high, xhigh, max), alongside modest text prompt overhead. Based on published platform pricing, developers observe standard generation requests landing between $0.025 and $0.035 each.

Here is how those figures calculate across four typical production workloads:

Workload Scenario	Input Configuration	Output Configuration	Cost Formula & Calculation	Estimated Request Cost
1. Text-to-Image Concept (Single 1024x1024 still)	200 text tokens (uncached)	~900 image output tokens	(200 × $0.000005) + (900 × $0.000030) = $0.000001 + $0.027000	$0.0270
2. Markup-Guided Edit (Source image + PNG overlay)	300 text tokens + 2 image inputs (approx. 1,600 image tokens uncached)	~900 image output tokens	(300 × $0.000005) + (1,600 × $0.000008) + (900 × $0.000030) = $0.0015 + $0.0128 + $0.0270	$0.0413
3. Iterative Edit Loop (Cached prompt + cached source image)	300 text tokens (cached) + 800 image tokens (cached)	~900 image output tokens	(300 × $0.00000125) + (800 × $0.000002) + (900 × $0.000030) = $0.000375 + $0.0016 + $0.0270	$0.0290
4. Animated GIF Loop (Multi-frame generation sequence)	500 text tokens (uncached)	~3,500 image output tokens (dense frame sequence)	(500 × $0.000005) + (3,500 × $0.000030) = $0.0025 + $0.1050	$0.1075

Notice that for complex operations like animated GIFs (which generate multi-frame sequences between 2 and 64 frames), output token consumption scales significantly. Production workflows utilizing prompt caching on fixed base layouts can shave up to 75% off recurring input costs during consecutive editing rounds.

Source: @Mr_Salio

How GPT Image 2.6 Pricing Compares

When evaluating GPT Image 2.6 against its direct predecessor and broader frontier multimodal models, pricing reflects an emphasis on feature expansion rather than per-token deflation.

For architectural comparison on OpenAI's text and reasoning ecosystem, see our analysis of GPT-6.1 Terra. Within the image generation line, the cost profile remains anchored to token counters established in the 2.5 series.

The table below contrasts GPT Image 2.6 with GPT Image 2.5 and comparative enterprise visual endpoints:

Model	Text Input (/1M)	Image Input (/1M)	Image Output (/1M)	Batch API Supported	Key Architectural Additions
GPT Image 2.6 (sunburst / flare)	$5.00	$8.00	$30.00	No (Launch restriction)	Transparent PNG Markup, SVG, GIFs, Complexity control
GPT Image 2.5 (sunburst / flare)	$5.00	$8.00	$30.00	Yes (50% discount)	Standard Inpainting, Mask support, Responses API
GPT Image 2 (Legacy Baseline)	Flat per-image rates	Flat per-image rates	Flat per-image rates	Yes	Standard generation (Retired legacy endpoint)
Gemini 4 Carbon (Multimodal Reference)	$2.00 (Promo)	$2.00 (Promo)	$10.00 (Promo)	Yes	1M Context Frontier Reasoning, Developer preview

The critical pricing reality separating version 2.6 from 2.5 is output token density. While the headline token price is unchanged at $30.00 per 1M image output tokens, developer analyses reported by third-party tracking portals indicate that GPT Image 2.6 emits up to 4.2 times more output tokens on low-quality runs and 3.3 times more on medium-quality runs compared to version 2.5. Consequently, low-tier calls that previously cost under $0.01 on version 2.5 may approach $0.025 to $0.030 under version 2.6 due to expanded token representations.

Free Tier, Limits, and Access

GPT Image 2.6 access is segmented across ChatGPT interfaces, developer environments, and direct API platform access.

+-----------------------------------------------------------------------------+
|                        GPT Image 2.6 Access Routes                          |
+------------------------------------+----------------------------------------+
| Consumer / Pro Interfaces          | Developer API Endpoints                |
| - ChatGPT Plus / Pro / Team / Edu  | - Direct OpenAI API (Metered Tokens)   |
| - ChatGPT Free (Rolling access)    | - Cloud Partner Workflows              |
| - Codex Integration                | - Hosted Gateways (kie.ai API Catalog) |
+------------------------------------+----------------------------------------+

Access Pathways

1. ChatGPT Web & Mobile: As documented in the ChatGPT Images 2.6 System Card, the model is available to ChatGPT users across web, iOS, and Android clients. ChatGPT Plus and Pro subscribers receive higher messaging quotas and priority generation, while free-tier users receive standard capped usage pools subject to peak server throttling.
2. OpenAI Developer Platform: API developers access the system via standard API keys targeting the model identifiers gpt-image-2.6-sunburst and gpt-image-2.6-flare. Usage bills directly against developer credit balances under standard organizational rate-limit tiers.
3. Hosted Third-Party Platforms: For developers orchestrating multimodal pipelines alongside other vision models, kie.ai offers GPT Image 2 within its unified inference catalog, allowing teams to benchmark earlier generational performance against updated models without maintaining disparate vendor billing accounts.
4. Third-Party Creative Apps: Platforms such as fal and Higgsfield integrated GPT Image 2.6 at release, with Higgsfield running promotional launch offers such as unlimited generation for seven days to new users on their consumer tier.

Rate Limits and Endpoint Constraints

Developers migrating from GPT Image 2.5 must account for endpoint changes:

• Endpoints Supported: POST /v1/images/generations and POST /v1/images/edits.
• Endpoints Excluded: Neither v1/responses nor v1/chat/completions tools accept GPT Image 2.6 models at launch.
• Batch Processing: The standard asynchronous Batch API is disabled for version 2.6. Workloads requiring bulk generation cannot currently leverage the standard 50% overnight batch discount.
• Mutual Exclusivity Constraint: In the POST /v1/images/edits endpoint, using the new markup parameter (a transparent PNG layer) strictly disallows providing a mask parameter in the same HTTP payload.

What We Don't Know Yet

Several enterprise operational parameters remain unconfirmed in public documentation:

• Per-Resolution Token Tables: OpenAI has not published an exact mathematical matrix documenting guaranteed output token consumption across non-standard aspect ratios (such as 1536x1024 or custom multiples of 16).
• Batch API Timeline: No official roadmap specifies when v1/batches support will be enabled to unlock 50% cost reductions for non-realtime rendering.
• Dedicated Hardware Commitments: Tier-5 enterprise commitments and throughput reservation pricing for high-concurrency 2.6 Flare instances have not been publicly disclosed.
• Long-Form GIF Pricing Ceilings: Maximum token caps for maximum-duration animated loops (up to 64 frames) remain unstated in the baseline documentation.

Frequently Asked Questions

Is GPT Image 2.6 free to use?

GPT Image 2.6 is not free in the API, where calls incur usage-based billing at $5 per 1M text input tokens, $8 per 1M image input tokens, and $30 per 1M image output tokens, but it is accessible to ChatGPT end users across standard subscription plans.

How much does the GPT Image 2.6 API cost per image?

GPT Image 2.6 API pricing charges per token rather than a flat per-image fee; typical single-image runs cost between $0.025 and $0.035 depending on input context, target resolution, and generated output tokens.

Does GPT Image 2.6 Flare cost less than Sunburst?

GPT Image 2.6 Flare and Sunburst share the exact same official OpenAI token pricing of $8 per 1M image input tokens and $30 per 1M image output tokens, though Flare generates faster.

Does the new Complexity parameter increase API cost?

The Complexity parameter (low, medium, high) carries no separate surcharge on API calls, though higher complexity may influence the volume of tokens generated by the model.

Is GPT Image 2.6 cheaper than GPT Image 2.5?

GPT Image 2.6 shares identical base token pricing with GPT Image 2.5 ($8/1M image input and $30/1M image output), but developer analyses report that version 2.6 can consume up to 3.3 to 4.2 times more output tokens on certain low and medium quality generation runs.

Are Batch API discounts available for GPT Image 2.6?

OpenAI lists the Batch API and Responses API endpoints as not supported for both GPT Image 2.6 Flare and Sunburst at launch, meaning 50% batch discounts cannot currently be applied.

---

What to Watch Next

Track whether OpenAI enables v1/batches support for GPT Image 2.6 in upcoming minor updates, monitor developer reports on exact GIF token expenditure across frame counts, and watch for independent ELO benchmarking updates on the Artificial Analysis leaderboard.

Try related models on Kie.ai: GPT Image 2.6, GPT Image 2 and GPT Image 2.5.