🎨 GPT Image 2 and Nano Banana image APIs are now live

Modelflare now supports two separate image-generation API contracts.

GPT Image 2 / OpenAI-compatible Images API

• Endpoint: POST /v1/images/generations
• Model: gpt-image-2
• medium / high quality
• Short-lived signed URL or b64_json
• Synchronous JSON or stream=true SSE
• Current pricing is about 50% of the official API price; actual billing depends on quality and size

The first GPT Image 2 release is text-to-image only, generates one image per request, and does not support edits, variations, or reference-image editing.

Native Gemini image API

• Nano Banana Pro: gemini-3-pro-image
• Nano Banana 2: gemini-3.1-flash-image
• Synchronous: POST /v1beta/models/{model}:generateContent
• Recommended for long requests: POST /v1beta/models/{model}:streamGenerateContent?alt=sse
• Nano Banana Pro estimated image prices: 1K / 2K US$0.06720 per image; 4K US$0.12000 per image
• Nano Banana 2 estimated image prices: 0.5K US$0.02241, 1K US$0.03360, 2K US$0.05040, 4K US$0.07560 per image
• Prices are estimates per image; the final charge may vary slightly with request content

Gemini image models use the native Gemini protocol and do not use /v1/images/generations. Send /v1beta requests through modelflare.dev.

Image Workbench can now switch directly between GPT Image 2, Nano Banana Pro, and Nano Banana 2.

Documentation and examples:

Modelflare grok-4.6 is now officially available 🎉

It supports both the Chat Completions API and Responses API, with the same base pricing as grok-4.5:

Input: US$2 per 1M tokens Output: US$6 per 1M tokens Cached input: US$0.5 per 1M tokens

To get started, simply set the model name to grok-4.6. grok-4.5 will remain available. Give it a try!

Modelflare has launched a limited-time Friend Referral Campaign 🎉

During the campaign, you can share your personal invite link. When a friend registers through your link and completes their first top-up, both you and your friend will receive Credits equal to 8% of the friend’s first top-up amount.

Campaign page: For ongoing commissions and payouts, you can apply to become an Affiliate Partner:

Subscription plans are live: get up to 20% more API credits for the same budget

ModelFlare 28-day subscription plans are now available across GPT, Claude, and Gemini model families.

US$99.99 → US$110 total US$1,499.99 → US$1,800 total

Credits are issued automatically in four installments, once every 7 days. Bigger plans unlock a higher bonus rate.

View subscription plans:

GPT-5.6 Luna Pricing Adjustment

To continue providing GPT-5.6 Luna reliably, we reviewed the cost of operating the service. Because the current price no longer covers network bandwidth, infrastructure, and operating costs, we have chosen to keep the model available rather than remove it, and have adjusted its base price to 50% of the price before the latest reduction.

New base prices:

• Input: $0.50 / 1M tokens
• Cached input: $0.05 / 1M tokens
• Cache write: $0.625 / 1M tokens
• Output: $3.00 / 1M tokens

The new price took effect on August 6, 2026. Existing discounts for each user group remain unchanged.

Thank you for your understanding and support.