New models, service status, and pricing changes are published here as soon as they happen.
44
Total announcements
2
Latest updates
Changelog
44 results
Due to Gemini risk controls, the following models are affected and are currently being fixed: Gemini series language models nano-banana-fast nano-banana-2-lite These models are currently under maintenance. We will post another notice once service is restored.
Gemini has fixed the issue internally. The following models that were under maintenance have now been restored: 1️⃣ Gemini series language models 2️⃣ nano-banana-fast 3️⃣ nano-banana-2-lite
The following models have been restored: nano-banana-pro nano-banana-2.1 nano-banana-2 Other models not mentioned in this notice have not yet been restored, and the recovery time is uncertain. Thank you for your patience — we are working to resolve this as soon as possible.
Due to a protocol change by Gemini, the following models are currently experiencing issues. We are working on a fix and will post another notice once service is restored! nano-banana-pro nano-banana-2.1 nano-banana-2 The cl and vip channels are currently working normally. If you need these models urgently, please use these two channels for now.
Due to strengthened risk controls by Gemini, the following models are affected and are currently being fixed: Gemini series language models nano-banana-fast nano-banana-2-lite These models are currently under maintenance. We will post another notice once service is restored.
The following models have been restored and are now available: gpt-image-2 gpt-image-2.5
An OpenAI algorithm update has affected the gpt-image-2 and gpt-image-2.5 models. We are currently working on it and will post another notice once service is restored!
New model: nano-banana-2.1 The integration method remains unchanged. Simply update the model name.
The long generation time issue with the gpt-image-2 and gpt-image-2.5 models has been fixed!
Due to changes in OpenAI's rules, generation times for the gpt-image-2 and gpt-image-2.5 models are currently longer than usual. Optimization is still in progress, and we will post another notice once it is complete. The following models have been placed under maintenance due to insufficient resources. We will post another notice once service is restored: gpt-image-2.5-sunburst gpt-image-2.5-flare
The following models have been restored: gpt-image-2.5-flare gpt-image-2.5-sunburst
The following models are currently under maintenance. We will post another notice once service is restored: gpt-image-2.5-flare gpt-image-2.5-sunburst
1️⃣ 480p: ¥0.05/second, supports 1–15 seconds 2️⃣ 768p: ¥0.07/second, supports 1–15 seconds 3️⃣ 1080p: ¥0.15/second, supports 1–10 seconds ⚠️ Supports reference images for video generation (up to 9 images) and reference audio (up to 3 files). Frequently Asked Questions: 1. How is video generation billed? Billing is based on resolution and duration. See the model list for details. 2. How long does video generation take? It depends on the configured duration and reference materials. Generation usually takes around 4–12 minutes and may take longer during peak hours. Thank you for your patience. 3. What should I do if the result is unsatisfactory? Video generation requires professional prompts. Overly brief prompts rarely produce high-quality results. Please improve your prompt and try again. 4. How long are generated files retained? Images and videos are retained for 2 hours. 5. Are credits refunded for failed or policy-violating requests? Yes. 6. When will ComfyUI be updated? We will post an announcement after the update. Thank you for your patience.
1️⃣ gpt-image-2.5-flare (2,000 credits | ¥0.10) 2️⃣ gpt-image-2.5-sunburst (2,400 credits | ¥0.12) ⚠️ Both models support transparent backgrounds and the quality parameter. See the documentation for parameter details.
1️⃣ gpt-image-2.5 (600 credits | ¥0.03) Supports 1K only 2️⃣ gpt-image-2.5-sunburst (3,000 credits | ¥0.15) Supports 1K, 2K, and 4K 3️⃣ gpt-image-2.5-flare (3,000 credits | ¥0.15) Supports 1K, 2K, and 4K ⚠️ Important FAQs: 1️⃣ Can I select image quality? Yes. Use the "quality" parameter. The default is "auto"; available options are "low", "medium", and "high". 2️⃣ How do I set 1K–4K resolution? As with gpt-image-2-vip, use the "aspectRatio" parameter and provide exact pixel dimensions. Ratios are not supported. Examples: 1:1 1K: "1024x1024" 1:1 2K: "2048x2048" 1:1 4K: "2880x2880" See the documentation below for details. 3️⃣ Which API and request format should I use? Both the old and new APIs are supported. Parameters are unchanged; use the same format as gpt-image-2 and replace the model name. 4️⃣ Are credits refunded for failed or policy-violating requests? Failed or policy-violating requests do not consume credits. 5️⃣ What is the difference between gpt-image-2.5-flare and gpt-image-2.5-sunburst, and which should I choose? See the official article below.
New models: 1. gpt-6-astra input: ¥4/M tokens output: ¥20/M tokens 2. gemini-3.5-flash-lite input: ¥0.3/M tokens output: ¥2.5/M tokens 3. gemini-3.7-flash input: ¥0.6/M tokens output: ¥3.5/M tokens 4. gemini-3.8-flash input: ¥0.6/M tokens output: ¥3.5/M tokens Deprecated model: 1. gpt-5.4 (discontinued by the official provider and no longer available)
1. The 2K and 4K resolutions for gpt-image-2-vip are back. Due to costs, the price will be adjusted to ¥0.10 per image for all resolutions. 2. GPT-5.6 models added gpt-5.6-sol input: ¥2.2/M tokens output: ¥13/M tokens gpt-5.6-terra input: ¥0.9/M tokens output: ¥5.2/M tokens These updates will take effect at 9:00 AM on August 12.
Maintenance for nano-banana-pro and nano-banana-2 has been completed. Both models are now available.
Due to changes in Google's policies, nano-banana-pro and nano-banana-2 are currently under maintenance. We will post another notice once maintenance is complete.
Due to a cloud provider maintenance upgrade, Grsai servers are expected to restart for approximately one hour. During the restart, all Grsai models and the website will be unavailable. The affected period is from 12:00 AM to 1:00 AM on August 8. To keep your service stable, please arrange a backup provider in advance.
nano-banana-pro and nano-banana-2 were unavailable due to an official Google update. The issue has now been fixed and both models can be used normally. Affected period: 03:30–08:28
An official OpenAI issue caused gpt-image-2 image generation failures this afternoon. Service has now been restored.
Due to an official OpenAI issue, gpt-image-2 is currently unable to generate images. The estimated recovery time is unknown. We will post another notice when the official service is restored.
Google has officially discontinued gemini-2.5-flash-image (Banana 1) and introduced gemini-3.1-flash-lite-image. Grsai API is making the following changes: 1. Model removed nano-banana 2. Underlying model changed nano-banana-fast now uses gemini-3.1-flash-lite-image instead of gemini-2.5-flash-image. 3. New model nano-banana-2-lite (same as nano-banana-fast) Key points: 1. Price and API usage remain unchanged nano-banana-fast remains ¥0.022, and its API usage is unchanged. Integrated users do not need to modify their code. 2. New model price nano-banana-2-lite: ¥0.022
📢 Second price adjustment for Banana CL and VIP channels We apologize for adjusting prices again. Recent risk-control measures have sharply increased costs. Despite a small adjustment a few days ago, these services are still operating at a loss. To maintain stable service, we must make a second price adjustment for the following models: nano-banana-pro-cl: 10,000 credits | ¥0.50 nano-banana-pro-vip: 10,000 credits | ¥0.50 nano-banana-pro-4k-vip: 18,000 credits | ¥0.90 nano-banana-2-cl: 6,000 credits | ¥0.30 nano-banana-2-2k-cl: 9,000 credits | ¥0.45 nano-banana-2-4k-cl: 13,000 credits | ¥0.65 ⚠️ Notes: 1. These changes will take effect at 12:00 PM on Monday. 2. If costs fall, we will lower prices accordingly. 3. CL/VIP channels are intended for enterprise users with sufficient budgets and very high requirements for generation speed and stability. 4. nano-banana-2 and nano-banana-pro are not affected. We appreciate your understanding and support.
CL channel model resolutions and prices will change at 7:00 PM tonight. Due to increased channel costs, supported resolutions and prices for CL models will be adjusted as follows: I. nano-banana-pro-cl Supported resolution: 1K only (2K and 4K removed) Credits: 6,000 Price: ¥0.30 💡 For 2K or 4K, use one of these upgraded versions: 1–2K: nano-banana-pro-vip (10,000 credits | ¥0.50) 4K: nano-banana-pro-4k-vip (16,000 credits | ¥0.80) II. nano-banana-2-cl Supported resolution: 1K only (2K removed) Credits: 1,600 Price: ¥0.08 💡 For 2K or 4K, use an upgraded version: New 2K version: nano-banana-2-2k-cl (4,000 credits | ¥0.20) 4K version price increase: nano-banana-2-4k-cl (6,000 credits | ¥0.30) Important: 1. These changes will take effect at 7:00 PM tonight. 2. nano-banana-2 and nano-banana-pro are not affected. 3. Failed CL and VIP Banana channel requests are refunded; policy-violating requests are not refunded.
An official update recently slowed down gpt-image-2 generation. The model has now been optimized and generation speed has returned to normal.
2K and 4K resolutions for the Banana models have been restored and are available again.
2K and 4K resolutions for gpt-image-2-vip have been restored and are available again.
An official Google bug is causing image generation errors at 2K and higher resolutions for the Banana model family. Please wait for Google to fix the issue. 1K generation is unaffected, and nano-banana-pro-vt is also unaffected.
A critical service update made the website unavailable from 2026-06-12 07:30:00 to 2026-06-12 09:30:00. Service has now been restored.
A network proxy line failure caused some models to be unavailable this afternoon. The issue has now been resolved and service has been restored.
New model: gemini-3.5-flash input: ¥1.2/M tokens output: ¥10/M tokens
A major Google algorithm update caused requests to nano-banana-2 and nano-banana-pro to fail today. The issue has been resolved and both models are now available.
1. Gemini now uses a more stable solution, and the following model changes have been made. 2. New models gemini-2.5-flash input: ¥0.3/M tokens output: ¥2/M tokens gemini-3-flash input: ¥0.4/M tokens output: ¥3/M tokens gemini-3.1-flash-lite input: ¥0.25/M tokens output: ¥1.5/M tokens 3. Price adjustments gemini-3-pro Previous: input: ¥1/M tokens output: ¥6/M tokens New: input: ¥1.5/M tokens output: ¥7/M tokens gemini-3.1-pro Previous: input: ¥1/M tokens output: ¥6/M tokens New: input: ¥1.5/M tokens output: ¥7/M tokens
Existing integrations do not need to switch. The new documentation adds a unified endpoint (https://{base_url}/v1/api/generate) that can call every model available through Grsai API. In the old documentation, Banana and GPT models use different endpoints. The new endpoint is more convenient, so choose whichever option best meets your needs. Important: Do not mix parameters from the old and new documentation.
Frequently Asked Questions: 1. Can I continue using the old documentation? Yes. It will remain valid permanently. 2. What is the difference between the old and new documentation? The new documentation is more beginner-friendly, includes stable one-click copyable examples that can be sent to AI assistants, uses one endpoint for all models, fine-tunes parameters, and provides examples in multiple programming languages.
Frequent OpenAI risk-control measures have continued to increase the operating cost of gpt-image-2-vip. After careful consideration, we have decided to adjust its price: gpt-image-2-vip: ¥0.045 → ¥0.065 We apologize for any inconvenience and appreciate your continued support.
New model: gpt-image-2-vip Starting price: ¥0.045/image Available resolutions: 1K, 2K, 4K Available quality levels: low, medium, high; default: auto Custom pixel dimensions: see the official documentation for details
Frequently Asked Questions: 1. Resolution and pixel-dimension options are being added to the batch generation tool (image.grsai.ai). 2. The following OpenAI-compatible endpoints are supported: /v1/chat/completions and /v1/images/generations.
Multiple network proxy providers went down simultaneously at 11:10 AM, making several models unavailable. Service was restored after two hours. Affected models: nano-banana nano-banana-fast nano-banana-pro nano-banana-2 gpt-image-2
The OpenAI-compatible completions API now supports gpt-image-2, allowing image generation directly through the chat endpoint.
1️⃣ New models added 2️⃣ Batch generation added (generate 1–12 images at once) 🌟 GitHub: https://github.com/31702160136/ComfyUI-GrsAI 🤖 Currently available models: gpt-image-2 nano-banana nano-banana-fast nano-banana-pro nano-banana-pro-cl nano-banana-pro-vip nano-banana-pro-4k-vip nano-banana-2 nano-banana-2-cl nano-banana-2-4k-cl
New model: gpt-image-2, priced at ¥0.03 per request.
The official provider has discontinued the following models: sora-image and gpt-image-1.5.
Gemini has recently reduced official compute capacity, causing continued instability on the CL and VIP channels. To ensure a better user experience, we switched to a new provider, which increased our costs. As a result, prices for models on these two channels will increase at midnight today. We apologize for the inconvenience and appreciate your support. Model | Previous price → New price nano-banana-pro-cl: ¥0.17 → ¥0.30 nano-banana-pro-vip: ¥0.35 → ¥0.50 nano-banana-pro-4k-vip: ¥0.43 → ¥0.80 nano-banana-2-cl: ¥0.06 → ¥0.08 nano-banana-2-4k-cl: ¥0.075 → ¥0.15